https://dx.doi.org/10.1016/j.autcon.2022.104192 This is the Pre-Published Version.

1 Artificial intelligence in green building

2 Caleb Debrah

3 [email protected]

4 Albert Ping Chuen Chan

5 [email protected]

6 *Amos Darko

7 [email protected]

8 Department of Building and Real Estate, The Hong Kong Polytechnic University,

9 Hong Kong

10 *Corresponding author

11 Abstract

12 The Architecture, Engineering and Construction (AEC) sector faces severe sustainability and

13 efficiency challenges. The application of artificial intelligence in green building (AI-in-GB) is

14 an effective solution to enhance the sustainability and efficiency of the sector. While studies

15 have been conducted in the AI-in-GB domain, an in-depth study on the state-of-the-art of AI-

16 in-GB research is hitherto lacking. To provide a better understanding of this underexplored

17 area, this study was initiated via a bibliometric-systematic analysis method. The study aims to

18 reveal the synthesis between AI and GB, as well as to highlight research trends along with

19 knowledge gaps that may be tackled in future AI-in-GB research. A quantitative bibliometric

20 analysis was conducted to objectively identify the major research hotspots, trends, knowledge

21 gaps and future research needs based on 383 research publications identified from Scopus. A

22 further qualitative systematic analysis was also conducted on 76 screened research publications

23 on AI-in-GB. Through this mixed-methods systematic review, knowledge gaps were identified,

24 and future research directions of AI-in-GB were proposed as follows: digital twins and AI of

25 things; blockchain; robotics and 4D printing; and legal, ethical, and moral responsibilities of

© 2022. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/.
26 AI-in-GB. This study adds to the GB knowledge domain by synthesizing the state-of-the-art of

27 AI-in-GB and revealing the research needs in this field to enhance the sustainability and

28 efficiency of the AEC sector.

29 Keywords: Artificial intelligence; Green building; Bibliometric analysis; Systematic analysis;

30 Sustainability.

2
31 1. Introduction

32 Buildings and construction account for the largest share of both the energy use (36%) and

33 carbon emissions (37%) in the world [1], making sustainability, including the efficient use of

34 resources, a severe challenge facing the Architecture, Engineering and Construction (AEC)

35 sector. To enhance the sustainability of the sector, there is a global trend of supporting and

36 promoting green building (GB). The US Environmental Protection Agency [2] defines GB as

37 “the practice of creating structures and using processes that are environmentally responsible

38 and resource-efficient throughout a building’s lifecycle”. According to the World Green

39 Building Council (WorldGBC) [3], GBs are buildings that, in their lifecycle, decrease or

40 eliminate damages to the climate and the environment, and enhance the quality of life of people.

41 GB, owing to its benefits, has received increasing attention from researchers and

42 practitioners worldwide, leading to increasing related empirical studies [4,5]. Along with this

43 is also a number of review studies [6–8]. Despite the usefulness of these review studies, they

44 have been based upon qualitative, manual analysis of the literature, which is prone to lack of

45 reproducibility, subjectivity and bias, and thus reduced reliability [9]. While recent review

46 studies attempted to address these limitations by adopting the quantitative bibliometric

47 approach [10,11], they also lack the in-depth understanding that a qualitative approach could

48 afford. To overcome these limitations while also enhancing the depth and breadth of

49 understanding, this study adopts the mixed-methods systematic review approach (see section

50 3) to review the artificial intelligence (AI) in GB (AI-in-GB) literature for the first time.

51 AI is defined as the science and technology of making intelligent machines that can reason,

52 learn, communicate, plan, move, operate objects, and solve problems [12]. AI has several

53 benefits such as enhancing productivity and communication [13]. Although previous studies

54 have applied AI-in-GB to enhance the sustainability and efficiency of the AEC sector [14,15],

55 no research has thoroughly and systematically reviewed the state-of-the-art in AI-in-GB

3
56 research to inform future research and improvements in practice. To fill this gap, this study

57 reviews current literature on AI-in-GB to identify research trends and gaps that can be tackled

58 in future research. The study seeks to answer the following research questions:

59 (1) What was the annual research publication trend of AI-in-GB from 2002-2021? The

60 literature search found the first AI-in-GB publication in 2002, as shown in section 4.1,

61 justifying the choice of 2002-2021.

62 (2) What are the influential AI-in-GB research journals and articles?

63 (3) What are the research hotspots of AI-in-GB?

64 (4) What are the future research needs of AI-in-GB?

65 2. Previous work

66 This section reviews existing reviews on GB and AI-in-AEC to identify the research gap,

67 justifying the need for this mixed-methods systematic review. A summary of previous reviews

68 is presented in Table 1 and discussed in the following sub-sections.

69 Table 1
70 Summary of reviews on GB and AI-in-AEC.
Research theme
SN Source Timespan Research method GB AI-in-AEC
1. Zuo and Zhao [6] Not specified Critical review ✓
2. Lu et al. [16] 1999-2016 Critical review ✓
3. Darko and Chan [7] 1990-2015 Critical analysis ✓
4. Darko et al. [10] 1974-2018 Scientometric review ✓
5. Zhao et al.[11] 2000-2016 Bibliometric review ✓
6. Bilal et al. [17] Not specified Literature review ✓
7. Faghihi et al. [18] 1985-2014 Literature review ✓
8. Irani and Kamal [19] 1990-2012 Systematic review ✓
9. Pan and Zhang [20] 1997-2020 Scientometric and qualitative analysis ✓
10. Darko et al. [21] 1974-2019 Scientometric analysis ✓
71
72 2.1. GB review

73 GB research has grown in recent years, with the rapid growth in practical GB

74 implementation, leading to a number of review studies exploring various facets of GB research.

75 One of the earlier studies that reviewed the growth and research development of GB is owed

76 to Zuo and Zhao [6], who found that GB studies can be classified into three categories:

4
 77 definitions and scope of GB; benefits and costs of GB; and ways to achieve GB. Darko and

78 Chan [7] analyzed GB research trend in construction journals from 1990-2015 and identified

79 that GB project delivery and developments has been a dominant topic. Other GB reviews

80 focused on building information modeling (BIM) for GB [16]. While aforementioned reviews

81 provide good knowledge of the GB body of knowledge, they are limited by their qualitative

82 approach which might introduce more subjectivity and bias, which might affect their reliability.

83 To overcome this limitation, recent reviews have employed the quantitative bibliometric

84 approach in reviewing the GB literature [10,11]. Despite their usefulness, none of the previous

85 reviews examined the application of AI-in-GB. The review that specifically analyzes AI-in-

86 GB is useful to researchers and practitioners in developing cutting-edge AI-enabled solutions

87 and technologies for application in GB projects to enhance both the sustainability and

88 efficiency of the AEC sector. Moreover, this study adopts the mixed-methods systematic

89 review approach to analyze AI-in-GB research for the first time.

90 2.2. AI-in-AEC review

91 The application of AI-in-AEC is leading to digital transformation in the industry while it

92 attracts the attention of researchers. Some qualitative reviews were conducted to understand

93 the knowledge and practice of applying AI-in-AEC. Irani and Kamal [19] reviewed and

94 identified the historical trends and current patterns in the use of intelligent systems in the AEC

95 industry. Bilal et al. [17] reviewed the application of big data technologies to construction.

96 Another review focused on automation in construction scheduling [18]. These qualitative

97 reviews, despite being prone to subjectivity and bias, had narrowed perspectives focusing on

98 either specific AI technologies or specific AEC application areas, e.g., construction scheduling.

99 To overcome the limitations of the qualitative reviews and offer a complete picture of the AI-

100 in-AEC research in general, Darko et al. [21] presented a quantitative scientometric analysis of

101 AI-in-AEC research. More recently, Pan and Zhang [20] also reviewed the roles of AI-in-AEC.

5
102 Although generic perspectives present limitations when applied to specific areas, no

103 previous review has focused on the application of AI to the specific area of GB. AI applications

104 to conventional construction might not be directly applicable to GB because GB projects differ,

105 especially in their quest to address higher sustainability and efficiency standards. Therefore, it

106 is worthwhile to specifically review and understand the application of AI-in-GB.

107 3. Research methodology

108 The aim of this study is to synthesize the domain knowledge and to identify the research

109 needs in the AI-in-GB space. To this end, this study deployed a “mixed-methods systematic

110 review”, which encompasses a “quantitative review (i.e., bibliometric approach)” and

111 “qualitative review (i.e., systematic approach)”. This method was chosen over the “mono-

112 method manual systematic review” because of its ability to reduce biased conclusions and

113 subjective judgments and interpretations, as well as “to enhance the depth and breadth of

114 understanding” [22].

115 The mixed-methods systematic review methodology combines both quantitative and

116 qualitative methodologies in a single research, for synthesizing and analyzing available

117 literature on a subject [23]. According to Pluye and Hong [24], the mixed-methods systematic

118 review is convergent in that it synchronously integrates different data and methods of analysis.

119 The mixed-methods systematic review has been widely used in previous studies [25]. In this

120 study, the mixed-methods systematic review was utilized to overcome the limitations and

121 capitalize on the strengths of both quantitative and qualitative methods when used in isolation.

122 By integrating the bibliometric analysis with the systematic analysis, the challenge of biased

123 and subjective judgement and interpretations could be addressed [23]. Bibliometric analysis

124 was selected as the quantitative method, whereas systematic analysis was selected as the

125 qualitative method. The two methods are described below.

6
126 The bibliometric approach refers to the mapping and visualization of large scientific dataset

127 [26], which is useful in studying and comprehending the structural and dynamic features of a

128 scientific domain [27]. Bibliometric analysis uses networks to represent how specific

129 disciplines, scientific domains, or research fields are conceptually, intellectually, and socially

130 structured [28]. In line, this study employed the bibliometric analysis to identify the knowledge

131 domains, research trends, and main research outlets regarding AI-in-GB.

132 Systematic analysis, instead, is effective in revealing knowledge gaps and suggesting areas

133 for future studies towards advancing knowledge [29]. Under these circumstances, the mixed-

134 methods systematic analysis has been developed to construct the full picture of AI-in-GB

135 knowledge while isolating key areas of AI-in-GB for in-depth analysis. Adopting this method

136 helps to triangulate and elaborate the study results [30].

137 This study involved four stages: search for publications (stage 1), application of exclusion

138 criteria (stage 2), bibliometric analysis (stage 3), and systematic analysis (stage 4). Based on

139 the results, knowledge gaps and promising future research directions were then presented. Fig.

140 1 provides an overview of the research methodology, details of which are discussed next.

7
141

142 Fig. 1. Research methodology overview.

143 3.1. Search for publications (stage 1)

144 As illustrated in Fig. 1, the first stage of this study involved the search for publications. A

145 data collection strategy was developed to retrieve the relevant data. Data collection of previous

146 literature is critical since it defines the knowledge area from which conclusions for the study

147 are drawn. Accordingly, the strategy and database for the literature search were selected

148 carefully. The keywords, combined with the Boolean operators “OR” and “AND”, presented

149 in Table 2 were used to retrieve the bibliographic data from Scopus. Scopus was used because

150 it has a higher indexing rate with a wider and more recent publications coverage than other

151 academic databases [31]. Besides, it has been largely acknowledged in past studies [21,25,32].

152 Additionally, it is critical to incorporate well-known keywords to enhance the validity and

153 reliability of data. Nonetheless, while this study aims to incorporate well-recognized keywords,

154 it is unfeasible to include all possible keywords in a single study [7]. Therefore, one could

8
155 expand this research in future to include keywords such as zero-energy buildings, fuzzy

156 clustering, K-Means, cognitive computing, sentiment analysis, swarm intelligence, and human-

157 machine interface. The Scopus searches were conducted using the search keywords on the title,

158 abstract, and keywords sections of publications with no limitations on date range, resulting in

159 a comprehensive dataset of 392 articles.

160 Table 2
161 Keywords and literature search results.
String Results
TITLE-ABS-KEY ( ( "Green building" OR "Green construction" OR "Green project" OR "Green 392
retrofit" OR "Green housing" OR "Green technology" OR "Sustainable building" OR "Sustainable
construction" OR "Sustainable housing" OR "Sustainable project" OR "High performance
building" OR "High performance construction" OR "High-performance building" OR "High-performance
construction" OR "High performance project" OR "High-performance
project" ) AND ( "AI" OR "Artificial intelligence" OR "Machine intelligence" OR "Machine
learning" OR "Expert systems" OR "Genetic algorithms" OR "Neural networks" OR "Case-based
reasoning" OR "Data mining" OR "Fuzzy logic" OR "Fuzzy sets" OR "Knowledge-based
systems" OR "Support vector machines" OR "Artificial general intelligence" OR "Computer
vision" OR "Deep learning" OR "Reinforcement learning" OR "Transfer
learning" OR "Algorithm" OR "Image recognition" OR "Natural language
processing" OR "NLP" OR "Supervised learning" OR "Unsupervised learning" OR "Pattern
recognition" OR "decision trees" OR "Random forest" OR "Robotics" OR
cognitive* OR automation* OR augment* ) ) AND ( LIMIT-TO ( DOCTYPE , "ar" ) ) AND ( LIMIT-
TO ( SUBJAREA , "ENGI" ) ) AND ( LIMIT-TO ( LANGUAGE , "English" ) )
Manual screening based on the results of AI-in-GB a 76
162 Note: The Scopus search was conducted in October 2021.
163 a The manual screening process and criteria are described in sections 3.4 and 5.

164 3.2. Exclusion criteria (stage 2)

165 The “document type” was limited to “article” and the “subject area” was limited to

166 “engineering”. Since the main aim of this study is to review literature on AI-in-GB, it was

167 necessary to filter out all papers outside the scope of the study. A brief review of the abstracts,

168 and in some cases, where the abstracts failed to provide sufficient information, the contents of

169 the initially identified studies, was therefore conducted. After filtering, 383 articles were found

170 to be relevant and considered valid for further analyses. The reason for limiting the study to

171 only articles was that, compared to other document types such as conference papers, articles

172 commonly have higher quality due to their relatively rigorous peer-review process. Moreover,

173 articles offer a more authoritative body of knowledge for bibliometric analysis [21,33].

174 3.3. Bibliometric analysis (stage 3)

9
175 To better understand the knowledge domains, multiple bibliometric software were utilized

176 to analyze the data. VOSviewer 1.6.17, CiteSpace 5.8.R3, and Gephi 0.9.2 software were used

177 for analyzing the data to develop and visualize the knowledge maps. The software were

178 selected to take advantage of their cooperative use. As Cobo et al. [28] noted, any robust

179 bibliometric study requires the synergetic use of alternative software for different types of

180 analyses. Many software tools for bibliometric analysis exist, with their strengths and

181 weaknesses identified in the literature [28]. VOSviewer, an easy-to-use software, provides

182 distance-based visualizations of bibliometric networks – indicating relatedness [26]. CiteSpace

183 is a Java application for analyzing and visualizing emerging trends in a body of knowledge and

184 their interrelatedness [34]. Gephi is an open-source software that can visualize all kinds of

185 networks [35]. The combined use of VOSviewer, CiteSpace, and Gephi allows data analysis at

186 higher quality [29]. Thus, in stage 3, VOSviewer, CiteSpace, and Gephi were used for

187 bibliometric analysis, forming the basis for the systematic analysis in stage 4.

188 3.4. Systematic analysis (stage 4)

189 A qualitative analysis of carefully selected papers (as illustrated in Fig. 1) was conducted,

190 following Harden and Thomas [23]’s proposal of mixed-methods systematic analysis based on

191 predefined criteria outlined in subsection 5.1. This stage was closely linked to the literature

192 search and bibliometric analysis in stage 3. That is, a targeted, comprehensive, and visual

193 examination of all the 383 papers analyzed in stage 3 was performed to identify those papers

194 that were highly relevant to the subject matter – “AI-in-GB”. Similar to previous studies

195 [25,29], the qualitative systematic analysis comprised comparing concepts, themes, theories,

196 developments and research focuses of the carefully selected papers. This was facilitated by a

197 thorough discussion to provide insights into AI-in-GB and provide directions for further

198 research. Section 5 presents detailed description of the systematic analysis process.

199 4. Bibliometric analysis

10
200 4.1. Annual publication trends

201 Fig. 2 shows the annual publication trend of AI-in-GB. It shows that the trend greatly varies

202 annually. The first attempt of AI-in-GB was a 2002 publication in the journal “Construction

203 Management and Economics”, wherein the authors developed fuzzy-set theory decision-

204 support model for sustainable housing indicators [36]. This implies that although the

205 application of AI-in-AEC began in 1970s [21], AI-in-GB only gained grounds after almost

206 three decades (2002). Thereafter, in the wake of recent increasing data availability and

207 computational capabilities, AI-in-AEC interest has grown [20] leading to its application in GB

208 at a higher level of intelligence. Since then, there has been a gradual growth in AI-in-GB

209 research with 2021 recording the highest number of publications (72) so far. The momentous

210 growth of AI-in-GB in the 21st century could be linked to the recent growing interest in “AI-

211 in-AEC” and “GB” research [10,20,21]. Howbeit, the number of publications is still

212 unsatisfactory and does not reflect the importance of AI-in-GB. Nonetheless, the growing

213 publication trend is promising, suggesting an increase of AI-in-GB. This trend is likely to

214 continue, as the application of AI-in-GB continues to receive support from research and

215 practice worldwide. From Fig. 2, it is evident that AI-in-GB is an emerging area with

216 significant scope for further research. Moreover, it is worth mentioning that the total number

217 of publications in 2021 is up to October, where the literature searches were conducted. The

218 figure might grow before the end of the year, as suggested by the increasing growth in annual

219 citations (green line). To support this claim, a logistic regression model was built with high

220 predictive power of 75%, accounting for at least 10% of the variation in publications. The

221 Hosmer and Lomeshow test (𝑥 2 = 13.23, p > 0.05) is considered robust with the values of Cox

222 and Snell R2, and Nagelkerke R2 being 0.114 and 0.162, respectively. However, the Omnibus

223 test result (𝑥 2 = 2.44, p > 0.05) was insignificant, indicating that the fit was not adequate for

224 the data [37]. More surprisingly, a negative effect was found between the time and likelihood

11
225 for publications to increase (31.1%). The mixed and inconsistent logistic regression analysis

226 results could be due to the small sample size (of 20 observations) which tend to produce

227 inconsistent estimates [38].

228 Under an adjusted R-square of 0.957 (95% confidence interval), a Gompertz function [20]

229 is employed to fit the predicted data, as shown by the red line in Fig. 2. When the fitted function

230 is implemented, the number of publications is expected to rise to above 80 by the end of 2021.

231 AI solutions that benefit GB are increasingly gaining attention in the quest to bring digital

232 innovations to GB, hence the growth in publications.

Number of publications
Gompertz fitting curve (95% confidence interval) - Adjsuted R-square: 0.954
Number of citations
90 1800
80 1600
70 1400
Number of publications

72

Number of citations
68
60 1200
50 1000
40 800
41
30 600
29 31
20 25 25 400
20
10 16 200
1 0 1 5 7 9 5 6 6 6 10
0 0

233 Year
234 Fig. 2. Number of publications from 2002 to October 2021.

235 4.2 Main research areas: keywords co-occurrence analysis

236 VOSviewer was used to develop keywords co-occurrence networks using author keywords.

237 Using author keywords for bibliometric analysis is widely recommended for identifying the

238 main research areas in a domain [25,29]. Keywords co-occurrence is the synchronous

239 occurrence of two keywords [39]. The VOSviewer generates distance-based maps, in which

12
240 the distance between two items indicates the relational strength, with a smaller distance

241 depicting a stronger relationship [26]. The item label sizes indicate the frequency of the terms

242 in relevant publications. Different colors represents different clusters developed by VOSviewer

243 [25]. Gephi was used to compute the degree centralities of the keywords, based on which they

244 are ranked. High degree centrality values represent prominent research areas. However, where

245 two or more keywords had the same degree centrality, the betweenness centrality metric (which

246 suggests prominent nodes for the highest values within the network) was used [10]. The results

247 of keywords co-occurrence analysis of AI-in-GB are presented next.

248 As noted earlier, AI-in-GB has received limited attention, as the Scopus search yielded only

249 383 relevant articles. From the 383 articles, a total of 1,380 keywords were found using

250 fractional counting. The “minimum number of occurrences” was set to 2, a threshold met by

251 163 keywords. VOSviewer thesaurus file function was used to merge similar terms (e.g.,

252 artificial neural network and ANN were merged into “artificial neural network”). However,

253 ‘green building’ and ‘green buildings’ keywords were not merged [10] due to their distinct use

254 in the literature to refer to the ‘construction process’ and the ‘product’, respectively. Similar

255 logic applies to other keywords. The resultant network comprised 130 nodes and 458 edges, as

256 shown in Fig. 3. Table 3 shows the top 50 keywords in Fig. 3 based on degree centrality.

13
257
258 Fig. 3. Main areas of AI-in-GB research (co-occurrence network of keywords).

259 Table 3
260 Top 50 AI-in-GB research areas.
Research areas Degree centrality Betweenness centrality Average year published Relative influence
Green building 54 1328 2017 1
Sustainability 29 89 2018 2
BIM 26 136 2018 3
Multi-objective optimization 24 173 2018 4
Data mining 22 140 2018 5
Sustainable construction 22 0 2017 6
Decision making 21 242 2015 7
Artificial neural network 20 54 2018 8
Machine learning 19 290 2020 9
Genetic algorithm 18 127 2015 10
Optimization 17 159 2018 11
Energy efficiency 17 103 2018 12
Building design 17 34 2015 13
Sustainable building 17 27 2016 14
Green buildings 15 72 2016 15
Building envelope 15 50 2018 16
Life cycle cost 14 106 2013 17
LEED 14 45 2016 18
Thermal comfort 13 36 2017 19
Renewable energy 12 109 2015 20
Energy consumption 11 35 2019 21
Life cycle assessment 10 90 2016 22
Energy simulation 10 63 2017 23
Linear regression 9 35 2016 24
Material selection 9 33 2017 25

14
 Fuzzy logic 9 28 2018 26
Sustainable development 9 0 2013 27
Sensitivity analysis 8 12 2018 28
Multi-criteria decision making 8 7 2017 29
Compressive strength 8 4 2021 30
Energy conservation 7 120 2018 31
Building energy performance 7 48 2015 32
Pattern recognition 7 16 2020 33
Indoor environmental quality 7 14 2018 34
Decision tree 7 7 2015 35
Support vector machine 7 7 2019 36
Multi-objective genetic algorithm 7 5 2010 37
Embodied energy 7 3 2019 38
NSGA-II 7 3 2019 39
Additive manufacturing 7 0 2020 40
Analytical hierarchy process 7 0 2014 41
Sustainable design 7 0 2019 42
HVAC 6 84 2018 43
Fuzzy sets 6 67 2016 44
Green technology 6 34 2016 45
Geopolymer concrete 6 9 2020 46
Knowledge discovery 6 2 2020 47
Environmental performance 6 0 2018 48
Built environment 6 0 2015 49
Building automation 6 0 2015 50
261
262 Several findings are discussed based on the rankings and how the research areas are related

263 as presented in Fig. 3 and Table 3:

264 (1) First, the “average year published” of the top AI-in-GB keywords ranges from 2013-2021.

265 (2) Second, certain research areas have gained increased attention, while other areas have been

266 under-studied. “Green building”, “sustainability”, “BIM”, “multi-objective optimization”,

267 “data mining”, “sustainable construction”, “decision making”, “artificial neural network

268 (ANN)”, “machine learning”, “genetic algorithm (GA)”, “optimization”, and “energy

269 efficiency” have been keen in AI-in-GB research. It may be argued that “machine learning”

270 application in GB has gained relatively more attention, with most applications using ANN

271 and GA. The research areas have therefore focused mainly on AI-methods (such as data

272 mining, machine learning, etc) and the application in GB (for decision making,

273 optimization, classification, energy efficiency, building design, material selection, etc). For

274 instance, Wang et al. [40,41] employed GA to optimize GB designs.

275 (3) To better appreciate the AI-methods applied in GB, Fig. 4 visualizes the top 34 most used

276 AI-methods. For example, it was discovered that, while decision support system (green

15
277 cluster) highly co-occurs with multi-criteria decision making, FAHP (fuzzy analytical

278 hierarchy process), AHP, DEMATEL (decision-making trial and evaluation laboratory),

279 Delphi method, fuzzy logic, and fuzzy sets; ANN highly co-occurs with linear regression,

280 and support vector machine (SVM) (violet cluster).

281
282 Fig. 4. AI-methods in GB.

283 4.3 Cluster analysis

284 Cluster analysis was conducted using CiteSpace to uncover the fundamental topics and

285 research hotspots in order to appreciate the structure of the AI-in-GB knowledge domain [33].

286 As illustrated in Fig. 5, filtering minor clusters resulted in a network of six large clusters

287 (identified by cluster IDs #0 to #5), as shown in Fig. 5. The cluster labels in Fig. 5 were

288 generated using the log-likelihood ratio (LLR) [34]. The large modularity Q of the network

289 (0.9198) suggests that clustering the network is particularly effective, with dense links between

290 nodes within clusters, and sparse links between nodes in different clusters [21]. Similarly, the

291 large mean silhouette value (0.9897) indicates that each cluster is highly homogeneous [21].

292 This implies that while few studies on AI-in-GB have been conducted, the available studies

293 embody a network with dense connection that addresses similar issues in the research area.

16
294
295 Fig. 5. AI-in-GB research clusters.

296 The six clusters are grouped into two types: AI-methods and GB applications (Table 4). All

297 the silhouette values are approaching homogeneity, confirming the earlier assertion that AI-in-

298 GB research is inward-looking, and has not benefited from adapting relevant ideas/theories

299 from other domains. The average duration over which a particular cluster has been researched

300 is shown by the mean (year). As shown in Table 4, the largest cluster (#1) has 22 members, a

301 silhouette value of 1.000 and is labelled as “using fuzzy logic” by LLR. Since the introduction

302 of fuzzy set theory by Zadeh [42] in 1965 to deal with uncertainty due to imprecision and

303 vagueness, it has been largely applied in many fields including GB. In current research, fuzzy

304 logic and fuzzy sets have been adopted for making multi-criteria decisions in GB projects, such

305 as decisions regarding risk assessment, performance assessment, and selection of GB materials

306 [43–45]. The second-largest cluster (#2) has 14 members, a silhouette value of 1.000 and is

307 labelled as “occupant comfort”. Since the advent of GB research in 1974, “occupant or thermal

308 comfort” has been one of the major research areas [10]. Recently, AI-methods such as GA and

309 reinforcement learning control have been applied to intelligently automate the provision of

310 occupant comfort in GBs [46–48].

311 Table 4
312 Summary of identified clusters
Type Cluster ID Size Silhouette value Mean (year) Cluster focus
AI-methods #1 22 1.000 2017 Using fuzzy logic

17
 #3 9 0.945 2012 Genetic algorithms
#5 11 1.000 2017 Multi-objective optimization
GB-applications #0 13 0.973 2012 Existing building
#2 14 1.000 2005 Occupant comfort
#4 13 1.000 2002 Green building design optimization
313
314 4.4. Citation burst analysis

315 Citation burst analysis, which is undertaken using CiteSpace, tracks the keywords that have

316 high frequency of occurrence over a certain period, such as topics with citations surges or fast-

317 growing topics [49]. A total of 33 keywords in the dataset experienced citation bursts. Fig. 6

318 presents the top 25 keywords with the highest citation burst. The year range for the reviewed

319 literature is represented by light green lines, whereas the duration of a citation burst event is

320 represented by a red line. Optimization (burst strength, 4.99; burst period, 2016-2018), decision

321 making (4.79; 2018-2019), design (4.6; 2011-2016), automation (4.51; 2005-2013) and energy

322 conservation (3.82; 2013-2016) were the top five keywords with the strongest burst. However,

323 the low burst strength ranging from 1.99 to 4.99 reinforces the need for more AI-in-GB

324 research.

18
325
326 Fig. 6. Top 25 keywords with the highest citation burst in AI-in-GB literature (2002-2021).

327 4.5 Most cited publications

328 In addition to the keywords analysis, citation information of the 383 articles was also

329 analyzed to reveal the top 20 most-cited publications on AI-in-GB (Appendix A). Knowledge

330 of the most-cited AI-in-GB publications informs researchers and practitioners on key

331 information sources. From top citation analysis, it is observed that AI has mostly contributed

332 to optimizing GB design, construction, and performance. Wang et al. [40,41], for example,

333 applied multi-objective GA to optimize GB design.

19
334 4.6. Top outlets for AI-in-GB research

335 Analyzing academic journals in any scientific domain is essential for readers and authors to

336 find the best sources of information and where to best publish, and for journal editors to make

337 relevant adjustment to their journals’ goals. Institutions and libraries may also benefit in

338 optimizing their investment in journals [21]. Table 5 presents the top research outlets for AI-

339 in-GB, developed using VOSviewer. The type of analysis was “citation”, and the unit of

340 analysis was “sources”. Additionally, the “minimum number of documents of a source” and

341 the “minimum number of citations of a source” were each set to 2. Selection of thresholds in

342 this study was based on past studies [21,25] and multiple experiments to generate the optimal

343 network. Using fractional counting, of 166 sources identified, 20 met the threshold. The

344 network comprised 20 nodes and 33 edges. Nodes and edges were resized based on their weight

345 strengths. Using Gephi (Fig. 7), the top research outlets were ranked based on weighted degree,

346 a widely used indicator for measuring the influence of nodes in controlling information across

347 networks [50]. The results (Table 5) show that Building and Environment (31), Energy and

348 Buildings (23), Journal of Cleaner Production (13), Applied Energy (12), and Automation in

349 Construction (7) were the top five outlets. These outlets may serve as reference points for

350 practitioners, researchers, and students on the state-of-the-art of AI-in-GB.

20
351
352 Fig. 7. Network of prominent outlets for research on AI-in-GB.

21
353 Table 5.
354 Top research outlets.
Outlets Number of Citations a Weighted Rank b
publications a degree value
Building and Environment 18 879 31 1
Energy and Buildings 34 1671 23 2
Journal of Cleaner Production 38 1155 13 3
Applied Energy 15 796 12 4
Automation in Construction 10 331 7 5
Energies 14 117 4 6
Journal of Construction Engineering and Management 7 296 4 7
Journal of Management in Engineering 4 186 4 8
Engineering, Construction and Architectural Management 2 7 4 9
Applied Thermal Engineering 3 124 3 10
Energy 6 135 3 11
Sustainable Cities and Society 12 303 2 12
Journal of Building Engineering 9 29 2 13
Journal of Architectural Engineering 4 15 2 14
Journal of Information Technology in Construction 2 41 2 15
Journal of Engineering, Design and Technology 3 5 2 16
Electronic Journal of Information Technology in Construction 2 48 1 17
Buildings 3 22 1 18
Building Simulation 4 17 1 19
Journal of Building Performance Simulation 2 9 1 20
355 a During the studied period (2002-October 2021).

356 b Ranking based on weighted degree values.

357 4.7. Scientific collaboration network analysis for AI-in-GB research

358 Scientific collaboration, also referred to as “co-authorship”, is necessary in any research

359 field to expedite access to funds, expertise, and specialties; limit research isolation; and

360 enhance productivity [33]. As such, the collaboration network analysis of influential

361 institutions and countries in AI-in-GB research is presented in the next sub-sections.

362 4.7.1 Influential countries

363 Fig. 8 depicts the global AI-in-GB research distribution by country in terms of number of

364 publications and citations. China and US emerged as the top contributors. However, it is

365 interesting to find that, although China contributed most in terms of number of publications,

366 the US received the highest number of citations. High citations numbers indicate the novelty

367 and significance of the underlying research, and the increasing importance governments attach

368 to it [51,52]. Countries such as Australia, Hong Kong, India, Malaysia, and UK have also made

369 good contributions. Nonetheless, there is considerable scope for increasing the number of

370 publications from most countries to improve global knowledge and practice on AI-in-GB.

22
 80 2237 Documents Citations 2500
70

Number of citations
Number of publications

2000
60
50 1345 1500
40
30 751 723 1000
614 609
20 374415353 406
286
238 176 238147 500
10 99 98 13373 99 78
73 63 28 27 25 25 25 17 17 16 15 14 11 10 10 9 9 7 6 6 23 5 56
5 5
0 0

Viet nam
Turkey
China

France
USA

India

Taiwan

Egypt

Brazil
Switzerland
Denmark
Saudi arabia
Germany
Singapore
UK

Spain

Iran
Hong Kong

Malaysia

Italy

South Korea
Australia

Canada
Countries

371
372 Fig. 8. Documents and country citation distribution.

373 A network was created with VOSviewer to provide a clearer picture of the research

374 contributions and the scientific collaborations of the countries. This analysis can help identify

375 countries that are highly engaged in the specific research field [49]. The type of analysis was

376 “co-authorship”, the unit of analysis was “countries”, and the counting method was “fractional

377 counting”. The “minimum number of documents of an organisation” and the “minimum

378 number of citations” were each set to 2 for optimal network. The criteria were met by 46 out

379 of 73 countries and the Gephi was used to visualize the resultant network (Fig. 9).

380 Countries that were more influential in the network were identified using the weighted

381 degree values [21]. Recoloring and resizing of nodes depended on the weighted degree values,

382 with larger nodes and darker colors signifying greater weighted values. Table 6 shows the top

383 30 most influential countries in the network.

384

23
385
386 Fig. 9. Collaboration network of influential countries in AI-in-GB research.
387 It is noticeable within this collaboration network that China, Australia, US, Malaysia, and

388 Hong Kong are the top five countries. China has the strongest collaboration with three of the

389 major contributors: US, Australia, and Hong Kong. With most (19%) of the AI-in-GB

390 publications emerging from China, such a strong relationship with other countries is expected.

391 Besides, China has increased research on GBs in recent years due to the high greenhouse gas

392 emissions [8]. On the contrary, US appears to have limited collaboration with other countries.

393 This observation could be attributed to the US government’s strategies to dominate the global

394 AI markets, and the mixed signals on the willingness to work with other countries to govern

395 AI [53]. Other strong relations identified within the network include China-UK, China-Taiwan,

396 Australia-Singapore, UK-India, Malaysia-Vietnam, and Malaysia-Saudi Arabia.

397 Comparatively, these nine strong collaboration relationships out of the 73 relations are very

398 limited. This can be associated with the restricted and/or a lack of comparative and cross-

24
399 country collaboration research. Generally, while developed countries exhibited strong network

400 collaborations, weaker relations were found within many developing countries. This underlines

401 the need for reforming policies to promote cross-country collaborations to advance AI-in-GB

402 research in terms of global collaboration, knowledge exchange, and enhanced productivity.

403 Table 6
404 Top 30 countries collaborating in AI-in-GB research.
Countries Number of publications a Weighted Degree value Relative influence
China 73 37 1
Australia 28 22 2
United States 63 18 3
Malaysia 25 13 4
Hong Kong 27 11 5
United Kingdom 25 11 6
Taiwan 16 8 7
Germany 10 8 8
India 25 6 9
Spain 17 6 10
Italy 17 6 10
Singapore 9 6 12
Canada 14 5 13
Iran 11 5 14
Vietnam 5 5 15
Saudi Arabia 5 5 15
South Korea 15 4 17
Egypt 10 4 18
France 7 4 19
Brazil 6 4 20
Nigeria 8 3 21
Switzerland 6 3 22
Pakistan 5 3 23
South Africa 4 3 24
Belgium 3 3 25
Chile 3 3 25
Finland 3 3 25
Austria 3 3 25
Denmark 5 2 29
Bangladesh 2 2 30
405 a During the studied period (2002-October 2021).

406 4.7.2 Influential institutions

407 Knowledge of institutional collaboration is critical to high investments and increased

408 interest in AI-in-GB research. Such a discovery is key to developing policies and building

409 lasting academic partnerships [54]. The type of analysis was “co-authorship”, the unit of

410 analysis was “organisations”, and the counting method was “fractional counting”. The

411 “minimum number of documents of an organisation” and the “minimum number of citations”

412 were each set to 2 for obtaining the optimal network. The resultant VOSviewer network

413 comprised 84 out of 574 organizations that met the criteria. The network (Fig. 10) visualized

25
414 with Gephi comprise 84 nodes and 71 edges. Nodes and edges sizes were resized and recolored

415 based on their weight strengths using the “hyperlink-induced topic search” in Gephi based on

416 their hub scores [21].

417
418 Fig. 10. Collaboration network of institutions in the literature on AI-in-GB.

419 Fig. 10 shows that only a few institutions in China, Hong Kong, Singapore, Taiwan, UK,

420 Australia, and South Africa have established AI-in-GB collaborative research relations.

421 However, the thickness of the edges suggest that these collaborations are not strong. To attain

422 the highest standard of academic research and discussion on AI-in-GB, institutions should

423 collaborate to benefit from varied knowledge and research experience as this is currently

424 lacking in body of knowledge. This is necessary to facilitate the sharing of knowledge and

425 ideas [21] on AI required in GB research and practice globally.

426 5. Systematic analysis

427 To provide in-depth analysis of research integrating AI-in-GB, a systematic analysis of

428 carefully selected articles is presented in this section. A qualitative screening and examination

429 of the 383 publications identified in stage 3 (Fig. 1) revealed 76 relevant articles, which were

430 included in the systematic analysis. The criteria for the selection of the articles are outlined as

431 follows:

26
432 5.1 Key article selection criteria for systematic analysis

433 Based on 383 AI-in-GB bibliometric records retrieved from Scopus, we adapted Kirchherr

434 and van Santen [55]’s and Antwi-Afari et al. [56]’s approaches to select the key articles for the

435 systematic analysis as outlined below:

436 (1) First, the 20 most cited articles were collected.

437 (2) Second, we selected five most recent works from the top five AI-in-GB journals based

438 on the weighted degree values: Building and Environment; Energy and Buildings;

439 Journal of Cleaner Production; Applied Energy; and Automation in Construction

440 (Table 6).

441 (3) Finally, we selected 55 articles at random based on their prominence in the literature

442 using forward-and-backward snowball manual search method [57]. This strategy begins

443 with identification of an initial set of papers (steps 1 and 2 above). Then, each study in

444 the start set was used to conduct backward-and-forward snowballing. In the backward

445 snowballing, relevant studies in the reference list of each study in the initial set were

446 identified. In the forward snowballing, the articles identified from the backward

447 snowballing formed a new start set which was subjected to another round of backward-

448 and-forward snowballing. This cyclical process was repeated until no new papers were

449 identified.

450 Titles or abstracts or full-text article, where the titles and/or abstracts were unclear, were

451 screened and were considered for inclusion if they were empirical studies on AI-in-GB. By

452 empirical studies, we refer to publications that were based on the description or assessment of

453 AI methods in providing a practical application in GB, for instance, on/off GB sites and

454 pre/post GB projects. After removing irrelevant articles and assessing the obtained articles on

455 the pre-defined inclusion criteria, 76 articles were found eligible for further analysis. Given the

456 systematic approach adopted, we believe that our sample is representative of the now-available

27
457 AI-in-GB literature. However, we do not claim to present an exhaustive overview of the AI-

458 in-GB literature. All articles examined are listed in Appendix A.

459 5.2 Key research areas of AI-in-GB

460 Since the introduction of AI-in-AEC, there has been growing research in areas such as

461 “knowledge representation and reasoning”, “information fusion”, “computer vision”, “natural

462 language processing” (NLP), “intelligent optimization”, and “process mining” [20,21]. Relying

463 on the wide AI-in-AEC approaches above, we classify AI-in-GB into four hot research areas,

464 as summarized in Fig. 11 and detailed below. More importantly, these research hotspots are

465 highly related to the clusters analysis (Table 4). For example, (1) “fuzzy rules and knowledge

466 discovery” is related to “using fuzzy logic”; (2) “big data and data mining” is related to

467 “existing building”; (3) “intelligent optimization” is related to “genetic algorithms” (GA), “GB

468 design optimization”, and “multi-objective optimization”; and (4) “building automation

469 system” is related to “occupant comfort”.

470
471 Fig. 11. Summary of main research areas.

472 5.2.1 Fuzzy rules and knowledge discovery

473 A foremost application of AI-in-GB is knowledge discovery and fuzzy rules. Fuzzy logic

474 evolved from the desire to train computer systems with human expertise [58]. The adoption of

28
475 fuzzy set theory (FST), introduced by Zadeh [42], has been applied to handle uncertainties in

476 GB and multi-criteria objectives in a fuzzy decision environment [59]. In recent years, there

477 has been various applications of the FST in GB such as the fuzzy synthetic evaluation (FSE),

478 fuzzy clustering, FAHP, Fuzzy analytical network process (FANP)-DEMATEL, and the fuzzy-

479 VIKOR-TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) as

480 presented in Table 7. For example, the “fuzzy clustering analysis” clusters objective things by

481 establishing fuzzy similarity relations according to their characteristics, degree of affinity, and

482 similarity [44].

483 Table 7.
484 Some studies for fuzzy rules
Method Purpose References
FSE GB risk assessment and management [45]
Fuzzy clustering GB multi-criteria performance assessment [44,60]
Assessing eco-building indicators [61]
FAHP GB multi-criteria performance assessment [58,62,63]
GB material selection [43,64]
Intuitionistic fuzzy sets Matching management of supply and demand of GB [65]
technologies
Fuzzy-VIKOR-TOPSIS GB risk management [66]
Health and safety risk assessment [67]
FANP-DEMATEL Ranking sustainability of GB material [68]
Ranking indicators for GB assessment [69]
Health and safety risk assessment [67]
Sustainable construction and demolition waste [70]
Fuzzy-Delphi method Sustainable construction and demolition waste [70]
Non-dominated fuzzy Green construction assessment [71]
decision support system
485
486 From Table 7, FSE have been applicable in many GB research areas to aid decision making.

487 Research has shown that integrating fuzzy logic with AHP is robust to deal with decision-

488 making problems with many alternatives [58]. AHP [72] is a structured technique for

489 organizing and analyzing complex decisions based on the hierarchy process. It therefore has

490 the ability to cope with the intuitive, the rational, and the irrational when making multi-

491 objective, multi-criterion, and multi-actor decisions. For example, the fuzzy-AHP has been

492 used to select GB materials [43,64] and to develop a green performance evaluation system for

493 construction site layout [62]. Additionally, Khoshnava et al. [68] combined the FANP and

494 DEMATEL to rank GB materials. The ANP, a generalization of AHP, applies a network

29
495 structure instead of a hierarchical structure [73]. DEMATEL, on the other hand, converts the

496 relationship among factors and fundamental dimensions from the complex system to a logical

497 organizational model [74]. Other multi-criteria decision-making (MCDM) techniques such as

498 TOPSIS [75] and VIKOR [76] have been combined with fuzzy prospect theory to solve

499 decision-making problems in GB [66]. It is evident that MCDM methods are based on the

500 trade-off between positive ideal solution and negative ideal solution – making it desirable in

501 decision-making [77].

502 5.2.2 Big data and data mining

503 Big data are usually high volume and high velocity data sets beyond the ability of traditional

504 databases to capture, manage, and process due to their high variety and sizes. Big data analytics

505 has the capacity to analyze such structured, semi-structured, and/or unstructured data using

506 advance AI [78]. Data mining, as a computation process, is therefore used to discover hidden

507 knowledge from large datasets and transform such knowledge into understandable structure for

508 future decisions [79,80]. GBs are now fitted with sensors (such as temperature sensors), power

509 and flow meters capable of providing big rich data streams by the minute, thanks to

510 advancement of smart metering and building automation technologies. This building data is

511 collected on a regular basis and can be examined to assist facility managers in improving

512 operational efficiency and reducing energy waste [81]. Recent developments in big data

513 analytics across the board [82] including GB have led to the development and application of

514 several data-driven machine learning algorithms such as artificial neural network (ANN),

515 convolutional neural network (CNN), K-Nearest Neighbor (KNN), multiple linear regression

516 (MLR), support vector machine (SVM) or regression (SVR), ensemble methods, association

517 rule mining (ARM), clustering analysis, and logistic regression. Over time, data mining has

518 been applicable in GB rating systems, energy consumption forecast, GB design modeling, and

519 GB cost and price prediction. This suggests that data mining is a critical decision-making

30
520 algorithm that eliminates the trial-and-error approach. Besides, due to the volume or big data

521 mined for analysis, clustering algorithms (such as hierarchical (connectivity-based), centroid-

522 based (k-means), distribution, and density-based clustering), are capable of dividing

523 multidimensional and heterogenous data into several clusters that are internally coherent and

524 externally separated [83]. Table 8 shows some of the relevant application of data mining in

525 GB.

526 Table 8.
527 Some studies for big data and data mining.
Method Purpose References
ANN GB cost and price prediction [84–86]
Construction schedule performance [85]
Building energy performance [44,87]
Determining influential factors of GB [88]
Predicting indoor environmental quality [89]
GB design optimization [90,91]
Predicting the strength of geopolymer concrete [92,93]
Analyzing and predicting the characteristics of wood [94]
CNN BIM-based GB design [95]
KNN BIM-LEED integration for GB design assessment [96]
Achieving LEED credits [97]
MLR GB cost and price prediction [86]
GB design optimization [91,98]
Achieving LEED credits [97]
Multivariate adaptive GB design optimization [98]
regression
Multi-polynomial and GB design optimization [91]
Stepwise regression model
Logistic regression GB cost and price prediction [85]
Locally weighted regression Predicting occupant energy consumption behavior [99]
SVM/SVR GB cost and price prediction [85]
Construction schedule performance [85]
GB design optimization [95,98]
Selection of target LEED for existing buildings (LEED-EB) [100]
Predicting occupant energy consumption behavior [99]
Ensemble methods GB cost and price prediction [85]
Construction schedule performance [85]
Environmental impact prediction [101]
Pattern recognition of GB markets [102]
Achieving LEED credits [80,100]
Predicting occupant energy consumption behavior [99]
Construction schedule performance [85]
NLP Collecting and classifying GB material information [103]
Assessing occupants’ satisfaction with LEED-certified buildings [104]
Attention and sentiment analysis of GBs [105]
Clustering analysis Mining thermal behavior of façade systems [83]
Pattern recognition of GB markets [102]
ARM Mining thermal behavior of façade systems [83]
Achieving LEED credits [106]
Naïve Bayes Predicting occupant energy consumption behavior [99]
CBR Support building green retrofit decision [107]
Achieving LEED credits [97]
528

31
529 Typically, the ANN has been used to predict the cost, price, and performance of GB [44,84–

530 86] based on historical data. ANN is a mathematical and computational model that attempts to

531 simulate a biological neural structure to imitate human learning process [108]. The back-

532 propagation NN (BPNN) is the most mature and widely used configuration of ANN which

533 conjoins a feedforward multi-layer perceptron with a BP algorithm [44,108]. The BPNN has

534 been used to predict the compressive strength of geopolymer concrete [93], energy

535 consumption [44], and the performance of GBs [85]. In recent years, ANNs such as the Radial

536 basis function neural network is adopted to improve the weighted influence non-linear gauge

537 system (RBF-WINGS) [88]. Other methods such as the SVM (a supervised learning method)

538 [109], and logistic regression (a common statistical method) can handle both classification and

539 regression problems [110]. Ensemble methods such as random forests, gaussian process

540 regression, extreme gradient boosting, gradient boosted, and AdaBoost decision trees combine

541 decisions from several weak models and vote for a final outcome [20]. For instance, Martínez-

542 Rocamora et al. [101], combined BIM-based life cycle analysis tools with the random forest to

543 predict the environmental impact of construction. BIM provides a collaborative platform to

544 help construction project to be digitally modeled in a virtual environment [96] and share

545 knowledge and information about a project to facilitate early decisions from cradle-to-grave

546 [111]. To automate GB design assessment, Jalaei et al. [96] demonstrated that a BIM-LEED

547 integration was feasible by using a Distance Weighted KNN (DWKNN). Apart from that, Wen

548 et al. [95] demonstrated that BIM is effective in different phases of a GB using CNN.

549 Other data mining techniques such as random forests, AdaBoost decision trees, and the

550 SVM have been useful in achieving LEED credits [80,100]. Recently, based on a NLP

551 approach, Guo et al. [104] assessed occupants’ satisfaction with LEED-certified residential

552 buildings using social media. Moreover, using solutions of past similar problems, data mining

553 can be employed to solve new problems [97]. A typical example is the use case-based reasoning

32
554 (CBR) approach to support green retrofit decisions [107]. The CBR process is usually

555 comprised of five parts being, represent, retrieve, reuse, revise, and retain [107]. Others such

556 as ontology with web crawling technologies can be used to collect and classify GB material

557 information automatically [103].

558 5.2.3 Intelligent optimization

559 Intelligent optimization lies in the ability to locate pareto optimal solutions for GB using AI

560 techniques over traditional approaches. A solution is said to be pareto optimal if it is not

561 dominated by any other solution in the performance space [40]. Although an optimization

562 problem can either be constrained to a single objective (to identify one optimal alternative) or

563 a multi-objective (simultaneously optimize multiple objective functions) [20], the multi-

564 objective optimization algorithm has been preferred in GB. This could be due to the

565 complexities and time constraints associated with the optimization challenges in GB. Hence,

566 available AI-methods which takes less time have been applied over the years in determining

567 optimal or near-optimal solutions for multi-objective GB constraints. Multi-objective

568 optimization algorithms, such as GA, evolutionary algorithms, and particle swarm optimization

569 (PSO), as outlined in Table 9 have been established as AI-in-GB. GA is an optimization

570 algorithm based on natural selection and population genetic mechanism. In a GA, every

571 possible solution is encoded as an individual, and all individuals form the population (i.e., all

572 possible solutions) [112]. Over time, variations of GA such as non-dominated sorting genetic

573 algorithm-II (NSGA-II), quantum GA, etc., have been developed.

574 Table 9.
575 Some studies for intelligent optimization.
Method Purpose References
Multi-objective GB design optimization [40,41,46,98,112–120]
optimization Life cycle costs and life cycle environmental impact [40,121]
assessment
Energy efficiency decision making and optimization [114,122,123]
Optimize indoor environmental quality [124]
BIM-based GB design and performance optimization [111,125]
GB retrofits [126,127]
Construction and demolition (C&D) waste transportation [128]
Optimize prefabricated buildings [129]
576
33
577 For instance, Yu et al. [46] developed a GA-BP multi-optimization model of NSGA-II to

578 assist GB designers in obtaining a set of optimal solutions to building designs. Similarly, Chen

579 et al. [124] incorporated a robust sensitivity analysis and the NSGA-II to optimize indoor

580 environmental quality in high-rise residential buildings. It has been found that an ANN model

581 can improve the performance of NSGA-II models in GB design optimization [116]. Besides, a

582 QGA has a faster convergence speed, stronger optimization ability, and can maintain

583 population diversity with better optimization results. Wang and Wei [112] demonstrated that

584 the QGA can simultaneously reduce the energy load and cost of building envelope by 35.3%.

585 As noted earlier, the BIM platform can aid construction professionals in GB decision-

586 making. For instance, Inyim et al. [117] integrated the NSGA-II and BIM to aid decision-

587 making during the design stage of construction projects. Likewise, a revised PSO algorithm

588 was applied to search for the trade-off between life cycle costs and life cycle carbon emissions

589 in a BIM platform [111]. Apart from the application of AI for new GB, the GA has also been

590 applicable in solving building retrofit optimization problems [126,127], optimizing C&D waste

591 transportation [128], and to optimize prefabricated buildings [129].

592 5.2.4 Building automation system

593 Building automation system, also referred to as smart-GB, is a classic example of modern

594 technologies that offers vast amounts of data on actual building operations, allowing buildings

595 to be monitored and controlled automatically and intelligently in real-time [79]. Using smart

596 technologies and metering system (such as temperature sensors, gas and fire sensors, power

597 and flow meters), GBs have the ability to generate huge amounts of data for facility

598 management [81]. For example, an automatic building energy management system is critical

599 in monitoring and managing efficient energy use to avoid wastage and reduce cost [130].

600 Building automation system therefore serves two key purposes in GB: point anomaly detection

601 and context anomaly detection. First, “point anomaly detection” builds consumption prediction

34
602 models using previous energy consumption time series. For example, energy consumption is

603 forecasted on a regular basis, and anomalies are detected by comparing actual deviation from

604 the expected value. On the other hand, “context anomaly detection” utilizes additional

605 information such as building and construction material, weather, etc., to define the anomalies

606 [131]. Several AI-methods are therefore included in building automation system – especially

607 the heating, ventilation, and air conditioning (HVAC) system – to develop automatic building

608 diagnostic tool for health monitoring, fault detection, and diagnostics [132]. This is necessary

609 to identify the ‘black holes’ of energy consumption caused by unforeseen human-related or

610 equipment defects such as faulty sensors, inefficient energy-use, etc. For example, the fuzzy

611 logic could be used to detect abnormal operating conditions and to generate fault signatures to

612 classify the fault type. ANN classification technique is then applied to fault signatures for

613 various fault types [132]. A list of some studies on building automation system is provided in

614 Table 10.

615 Table 10.

616 Some studies for building automation system.
Method Purpose References
Expert system Control and monitor outdoor lighting control system; perform [133]
load estimate and fault diagnosis
Machine learning Fault detection and diagnosis of chillers [134]
model Predicting the efficiency of GB natural ventilation systems [48,135]
Point and context anomaly detection of energy consumption [131]
Building energy optimization [79,136,137]
Deep learning Predicting the efficiency of GB natural ventilation systems [135]
model Estimating annual heating demand [138]
Predicting building energy consumption [139]
617
618 Yan et al. [134] observed a higher prediction accuracy and lower false alarm rates for fault

619 detection and diagnosis using machine learning techniques. In using the ANN and PSO,

620 Gonçalves et al. [137] implemented a smart energy management system that can be applied to

621 both new and existing buildings and with any level of HVAC technology. Likewise, to predict

622 the efficiency of natural ventilation systems for GBs, Park and Park [135] adopted machine

623 learning and deep learning models to measure indoor and outdoor environmental variables.

624 Similarly, Chen et al. [48] used reinforcement learning to optimize HVAC and window systems

35
625 for natural ventilation. Westermann et al. [138] used a deep temporal CNN in estimating annual

626 heating demand based on multivariate weather data. Besides, the simplest form of AI, expert

627 system [133] is also capable of performing load estimate and fault diagnosis in building.

628 Additionally, it has the ability to effectively monitor and control lighting system in real-time.

629 Expert system uses human knowledge to solve problem that normally would require human

630 intelligence [58].

631 5.3 Methodological characteristics of AI-in-GB

632 GB is an important measure in dealing with energy and environmental problems in the

633 construction industry [140] which is riddled with difficult and complicated challenges [21]. AI

634 has therefore evolved as a powerful tool in solving such complex challenges [21] present in

635 GB. As discussed earlier, AI-in-GB can be categorized into four major research areas: fuzzy

636 rules and knowledge discovery; big data and data mining; intelligent optimization; and building

637 automation system. AI-in-GB has been useful in augmenting or automating the decision-

638 making process through prediction, optimization, digitalization, risk management, and

639 construction health monitoring and evaluation. To achieve these, a number of AI-methods have

640 been used to handle the big data in GB acquired through several techniques such as simulations

641 and experimental tests, questionnaire and expert surveys, sensor-based technologies, wearable

642 technologies, smart metering, and GB-related databases and websites such as the USGBC

643 website. Table 11 provides an overview of the type of datasets required in the identified

644 research hotspots in AI-in-GB. The datasets and sample sizes employed in the different AI

645 methods applicable to GB are discussed as follows:

36
646 Table 11.
647 Summary of methodological characteristics, strengths, and limitations of AI-in-GB.
Research area Dataset Strengths Limitations References
Fuzzy rules and - Questionnaire survey - Avoid likely risks blindly in - No criterion on the sample size [43,45,58,62,64–68,70,71]
knowledge discovery - Interview decision-making - Small sample size and sampling biases
- Expert panel - Deal with multi-criteria decision- - Absence of reliable databases
- BIM-based environmental data making problems and uncertainties - Inaccurate simulation results
Big data and data - Statistical data on GBs and green - Increased efficiency - Incomplete or missing data [44,79,80,83–88,90–92,94–
mining finance - Cost and time savings - Data unavailability 97,99,101–106]
- Historical energy consumption data - Reliability and improved accuracy - Poor data quality or incorrect data or noisy data
- Simulated energy consumption data - Simplicity - Lack of effective and convenient tools to perform
- Real-time energy consumption data - Learning from limited datasets the large dataset analysis
- BIM models and data - Large datasets - Small sample size and sampling biases
- GB certification data - Under-reporting and under-coverage biases
- GB material data - Black-box problem
- Questionnaire survey - Internet challenges
- Inaccurate simulation results
- Over-fitting problem
Intelligent - BIM data of GB materials (quantities, - Increased efficiency - Inaccurate simulation results [40,41,46,112,116,119,126–
optimization cost, and sustainability data) - Cost and time savings - Data unavailability 129]
- GB design data - Stronger and better optimization
- Life cycle assessment data (natural ability
resource extraction, and building - Can maintain the population
material production, on-site diversity
construction, and transportation) - Generalization ability
- Prefabricated buildings - Reliability and improved accuracy
- Historical building load profiles
(e.g., energy, water)
Building automation - Historical building load profiles - Best for prediction - Limited training sample size [79,131,135–139]
system (e.g., energy, water) - Cost and time savings - Training error
- Real-time energy consumption data - Robust and reliable - Inaccurate simulation results
- Environment data (temperature - Increased efficiency - Data unavailability
statistics, humidity statistics, weather, - Reliability and improved accuracy - Incomplete or missing data
and holiday information) - Poor data quality or incorrect data or noisy data
- Lack of effective and convenient tools to perform
the large data set analysis
- Equipment failure or human operation errors
648

37
649 • Sample-size effects in research is very critical since it can easily contaminate the design

650 and evaluation of a proposed system [141]. However, the issue of the appropriate

651 sample size especially for AI algorithms remains unclear and are largely unreported in

652 the literature [142,143]. In a recent review on the sample-size determination for

653 machine learning algorithms, it was discovered that the sample sizes ranged from two

654 to 90,000 per feature or attribute. However, there are no generally acceptable methods

655 for calculating the required sample size for a given model [143]. Besides, since

656 sampling cannot be done is isolation, there are no special right decision for determining

657 sample size for a research [144]. That said, the sample-size determination methods,

658 training and testing percentages, number of inputs and outputs, feature selection, and

659 error estimation where necessary for an optimum performance of a model is outside the

660 scope of this study.

661 • In MCDM techniques, – such as fuzzy logic, fuzzy sets, AHP, and DEMATEL –

662 competence, qualification, and experience are more important than sample size when

663 choosing experts [145]. This is because fuzzy rules and knowledge discovery rely on

664 human expertise to train computer systems to solve problems hence requires human

665 intelligence. Therefore, the accuracy is dependent on expert knowledge and experience

666 [58]. Hallowel and Gambatese [146] defined an expert as a construction engineering

667 and management graduate and professional with at least five years hands-on experience

668 peculiar to the construction site. For expert opinions, five respondents are considered

669 adequate [147]. Respondents in the questionnaire surveys, interviews, or expert panels

670 were construction professionals (such as engineers, project managers, architects,

671 building designers), contractors, suppliers, and government representatives. Sample

672 sizes ranged from seven [64,67] to 120 respondents [58]. Other studies used two

673 separate five-member expert panels in three different rounds [65]. A few studies [44]

38
674 used building historical data to simulate the actual operation to develop a prediction

675 model based on fuzzy clustering using a total of 8,226 groups of data.

676 • Data mining algorithms relate to the analysis of available big data such as GB historical

677 data [44], real-time data [87], and statistical data [88]. The datasets are obtained from

678 certified projects (e.g., LEED) [80,84,96,97,102,104,106], green finance and GB

679 databases [88], historical and real-time building load profiles [44,87], and questionnaire

680 surveys [95]. Despite the characteristic large volumes of ‘big data’, data mining

681 algorithms can handle both large [44] and very small datasets [87]. For instance, the

682 ANN was used to analyze GB and green finance data retrieved from 21 major banks in

683 China [88] and 76 LEED certified projects [96]. Besides, a survey was conducted with

684 16 LEED managers’ to obtain their opinions based on certified projects [106].

685 Moreover, Wen et al. [95] used a questionnaire survey of 2000 respondents in a CNN

686 model to assess the effectiveness of BIM in GB design. It is observed that data mining

687 algorithms have also been used to analyze large sample sizes of building data such as

688 8,226 energy consumption data [44], 16,761 online GB reviews [104] and 21,000

689 simulation data points from buildings with conventional panels and ultra-high-

690 performance fiber-reinforced-concrete façade panel. Nonetheless, Wang et al. [116]

691 claim that ANN has the ability to learn from limited datasets.

692 • Intelligent optimization algorithms similarly can handle both small and large sample

693 sizes. For instance, NSGA-II models can handle both small datasets (30 cases) [120]

694 and large datasets (5,610 cases) [98]. It is observed that BIM-based data and

695 experimental simulation data is accepted and widely used developing GB multi-

696 objective optimization models. Other datasets are lifecycle assessment data including

697 natural resource extraction, GB material production, on-site construction, C&D

39
698 transportation distance [110,126], building load profiles [126,127], and prefabricated

699 buildings [129] have been useful in optimizing GB designs and green retrofits.

700 • Building automation system seeks to operate GB intelligently and in real-time. This is

701 possible through real-time data collected with smart technologies such as sensor-based

702 technologies, wearable technologies, and smart metering systems which have the ability

703 to generate huge amounts of data for facility management [81]. Advances in expert

704 systems, machine learning, and deep learning algorithms can handle both small and

705 large sample datasets. To characterize building load profiles, for instance, 144 daily

706 samples of smart metering data were obtained [131]. On the other hand, Westermann

707 et al. [138] used deep temporal convolutional networks to process 150,000 annual

708 hourly weather time series data. Gonçalves et al. [137] also used a large sample (35,040

709 cases) provided by EnergyPlus to develop an adaptable systems of intelligent

710 supervisory predictive control for buildings. The findings suggest that the building

711 automation system can handle both large and small sample sizes to understand the

712 behavior of GBs for real-time intelligent operation, monitoring, and evaluation.

713 • Due to the high volume, high velocity, high variety, and sizes of datasets in GB, they

714 are beyond the ability of traditional databases to capture, manage, and process [78]. The

715 development of software applications in building design processes have grown over the

716 years to handle complex data in GB. For example, Python has been widely used in

717 analyzing statistical data [88], GB material data [103] and BIM data. Other software

718 such as MATLAB [44,63,92,96,116], Statistica [84,94] and the R software [98,100]

719 have also seen extensive use in past studies. MATLAB [148], for example, is a high-

720 performance programming software for engineering calculation, which can perform

721 numerical analysis, matrix calculation, scientific data visualization, modeling, and

722 simulation of nonlinear dynamic systems. Besides, the Sina Weibo platform [105] with

40
723 web crawler technology [103,105] have been used in text mining methods. On the other

724 hand, simulation software such as TRNSYS [44,120,129], EnergyPlus

725 [87,98,99,113,138], jEPlus [124,129], Energy_10, Green Building Studio web tool

726 [87,96,125], and eQuest [87] have been used in past studies. TRNSYS (Transient

727 system simulation tool), for instance, can communicate with MATLAB through

728 communication object module interface [44]. Besides, while EnergyPlus and eQuest

729 are more suitable for the final design stages where a good level of detail is provided,

730 Green Building studio, which uses DOE-2 as the simulation engine, is fit for any design

731 stage [87]. Additionally, the Monte Carlo simulation, a powerful statistical tool, has

732 been used to forecast lifecycle costs of GB materials [121]. Finally, MCDM methods

733 such as ANP and AHP can be implemented in software such as Super Decision [68]

734 and Expert Choice 2000 [58].

735 • Notably, case studies were used to validate the developed AI-in-GB models or results

736 in majority of the studies.

737 5.4 Strengths and limitations of AI-in-GB

738 As presented in Table 11, there are numerous benefits that comes with the application of

739 AI-in-GB. Research has shown that AI-in-GB has been successful in reducing carbon

740 emissions to promote sustainability at increased profitability and reduced cost through efficient

741 approaches such as resource and waste optimization, risk management, monitoring and

742 evaluation, and automation. As shown in Table 11, similar strengths across the board include

743 increased efficiency, cost and time savings, and reliability and improved accuracy. Others such

744 as increased productivity, simplicity, reduced mistakes and omissions, faster prediction, and

745 resource optimization have been widely discussed in the literature [149].

746 While acknowledging the strengths in AI-in-GB, numerous challenges persist. An overview

747 of the limitations of AI is presented in Table 11. Data unavailability, incomplete data, poor data

41
748 quality, absence of reliable databases, inaccurate simulation results, small sample size, and

749 sampling biases were common limitations to AI-in-GB. Other limitations such as data

750 complexity, high initial costs, data privacy and security issues, ethical and legal issues, cultural

751 challenges, lack of expertise, and the “black-box” problem have been discussed in the literature

752 [149–151]. Machine learning systems, for example, use a black-box approach, which implies

753 that they do not explain the ‘why’ of their conclusions. It is therefore critical to employ

754 explainable AI to create explainable models that allow humans to understand, trust, and control

755 the systems [149,151]. Moreover, issues relating to the impact of culture, personal and religious

756 values on the acceptance and adoption of AI [150] should be investigated further.

757 Additionally, as noted earlier, issues regarding sample sizes and data quality have been

758 discussed in the literature. This is because data quantity and quality impact the accuracy of

759 models [79,135]. Hence, data collected for analysis usually undergoes “data preparation” or

760 “data pre-processing”. Data pre-processing helps to enhance data quality by removing outliers,

761 inconsistencies [83], and imbalances [85]. This comprises “data cleaning”, “data

762 transformation”, and “data reduction” before they are used to train prediction models [79].

763 Besides, for robustness and validation purposes, most AI algorithms splits datasets for

764 “training” and “testing” purposes – hence the need for large datasets. In such instances, to

765 eliminate the issue of overfitting and minimize bias associated with random sampling of the

766 training and testing data samples, especially when comparing the predictive accuracy of two or

767 more research methods, researchers often use cross validation to minimize the bias [100]. To

768 balance the class distribution in a dataset, the synthetic minority oversampling technique

769 proposed by Chawla et al. [152] is employed. Occasionally, where setting aside a part of a

770 small dataset for validation purposes is unaffordable, and it must all be consumed by the

771 training dataset, the “out-of-bag’ score provides a good trade-off. This provides insights on the

772 how the model could have behaved with a larger dataset that would have allowed splitting into

42
773 training and testing sets [101]. Moreover, to tackle the challenges associated with many

774 irrelevant and redundant variables and often comparably few training samples, variable and

775 feature selection have become the focus of many research [153]. They are essentially divided

776 into wrappers, filters, and embedded methods [102,153]. Finally, where the data gap is

777 extremely large, data normalization can reduce model prediction errors, improve convergence

778 speed, and model training efficiency [139,154].

779 6. Discussion and future research directions

780 This study has explored the state-of-the-art in AI-in-GB through a bibliometric-systematic

781 analysis. Unlike previous studies, the bibliometric-systematic analysis of AI-in-GB was

782 employed to: (1) synthesize the full picture of the research area, and (2) reveal the gaps, and

783 research needs thus justifying the need for the present research. To do this, bibliometric

784 techniques with the aid of multiple software such as VOSviewer, CiteSpace, and Gephi were

785 used to understand the key research areas, research outlets and scientific collaboration analysis

786 of countries and institutions in AI-in-GB. A further qualitative-systematic analysis was

787 conducted on key selected relevant articles on AI-in-GB.

788 So far, this study has identified the annual publication trend of AI-in-GB which reveals an

789 increasing and promising research area. A keyword co-occurrence analysis with VOSviewer

790 showed that GB, sustainability, BIM, multi-objective optimization, data mining, sustainable

791 construction, ANN, and GA have received special attention in the literature. A CiteSpace

792 cluster analysis identified six clusters which were mainly focused on AI-methods and GB

793 application. A more in-depth systematic analysis shows that past studies have focused on four

794 research hotspots: fuzzy rules and knowledge discovery, big data and data mining, intelligent

795 optimization, and building automation system. Finally, we evaluated the methodological

796 characteristics, strengths, and limitations of AI-in-GB.

43
797 It is therefore critical to identify and explore relevant directions for future research in order

798 to strengthen this research area. As presented in Fig. 12, there are key opportunities in four

799 major areas applicable in AI-in-GB. This suggest that more advanced technologies inspired by

800 AI will play a major role in digitizing, augmenting, and automating GB to promote the

801 efficiency and sustainability of the AEC sector. As discussed below, the application of AI-in-

802 GB, though a burgeoning research area, provides good grounds for future research.

803
804 Fig. 12. Future directions of AI-in-GB.

805 6.1. Digital twins and AIoT

806 Digital twins, along with AI of things (AIoT), data mining, and machine learning

807 opportunities [155] can offer great potential in the transformation of today’s GB. The digital

808 twin is a realization of the cyber-physical system for visualization, modeling, simulation,

809 analyzing, predicting, and optimizing which incorporates three components: the physical

810 entity, virtual entity, and the connection of data to form a practical loop [155]. On the other

811 hand, AIoT is the new generation of internet of things (IoT), which incorporates AI-methods

44
812 into IoT infrastructure for more efficient IoT operation and data analysis. IoT is therefore a

813 network of interconnected physical devices such as wearable and mobile devices, sensors,

814 wireless technologies (e.g., RFID), 3D laser scanners, sensors, actuators, and drones which is

815 attached to a construction resource (e.g., GB) to gather real-time data about the state of

816 operations of the project [20,149]. Since AIoT is powered by AI, it excels in the synthesis and

817 interpretation of data acquired in high volumes and velocity via IoT infrastructure, which can

818 then be transported to the virtual world for additional analysis via the digital twin technology.

819 In the virtual model, simulation, prediction, and optimization are conducted by learning data

820 from numerous sources which can provide instant solutions to direct real-world processes and

821 make them adapt to changing environments [20,149]. With the growing need for green retrofits

822 and investments [8], the cooperative use of digital twins and AIoT is very timely. With the aid

823 of laser scanners, BIM models of existing buildings could be developed and delivered to a

824 virtual world for additional analysis. The web-based integration of AIoT captures significant

825 amounts of data to enrich the BIM for green retrofit decisions, and BIM as a digital

826 representation can be the beginning point of the digital twin. Pan and Zhang [20] agree that the

827 convergence of AIoT, and BIM under 5G wireless communications would become hotspots

828 for future research. More so, a more complex approach with the inclusion of digital twins can

829 greatly improve the efficiency of the data gathering, transporting, and processing using cloud

830 computing methods towards smart-GB. Significant advances in cloud virtual and augment

831 reality [20,25] could also be explored in the proposed integration of BIM, AIoT, and digital

832 twin under 5G networks for a better comprehension of the complexities and interdependencies

833 during the construction of both new GB and green retrofits. Besides, future research could also

834 investigate the possibility of integrating the proposed technology with existing GB ratings

835 systems such as LEED (US), BREAAM (UK), and BEAM Plus (Hong Kong), and Green Star

836 (Australia), etc., to facilitate green certification processes.

45
837 6.2. Blockchain

838 Since 2008, blockchain has attracted significant interest in both financial and non-financial

839 applications due to the immutability, transparency, auditability, security, trustworthiness, and

840 fast-nature of its transactions [156,157]. A blockchain is essentially a distributed database of

841 records, or a public ledger of all transactions or digital events that have been executed and

842 shared among participating parties [157]. Blockchains can either be un-permissioned (allowing

843 public access) or permissioned (private access). With such abilities, blockchains are being

844 explored in various sectors including the construction sector. Some key opportunities for the

845 adoption of the blockchain technology in construction include smart energy, cities, homes,

846 organizations, transportation, BIM, and construction management and business [158].

847 Nonetheless, the now-available limited literature shows that the application of blockchain in

848 the construction sector is very limited [158] with opportunities for increased growth. Due to

849 the immutability, transparency, and traceability of data, blockchain can be used to promote GB

850 development to ensure resource optimization, reduced carbon emissions, and energy efficiency

851 of GB. For instance, green certification and verification could be digitized via blockchain and

852 IoT chips embedded in GBs [8]. Further, sensors tagged into the blockchain through IoT chips

853 could form the basis for assessing the performance of GBs to inform investors of the overall

854 environmental impact of developers in accessing green finance for GB. Besides, these data can

855 be used by developers to access incentives such as tax rebates, tax holidays, green insurance,

856 and green credits [8]. Moreover, the decentralized feature of blockchain can be integrated with

857 BIM to collect large amounts of data from the various stages of GB projects and share data

858 among stakeholders for life-cycle cost and environmental assessment. Such an integration

859 would allow the BIM model to be regularly updated to ensure an automated and streamlined

860 project delivery to promote productivity, trustworthiness, reduced cost, transparency, security

861 of data, effective collaboration and stakeholder management, and supply chain management

46
862 and resilience [20,157]. Further, smart property via blockchain using smart contracts could be

863 explored in GB projects to deal with the challenges associated with GB certification costs,

864 construction cost, payment fraud, and GB ownership [20,157]. With smart contracts, all GB

865 transactions will be accessible to all parties in real-time.

866 6.3. Robotics and 4D printing

867 Robots are intelligent machines that perform physical tasks in everyday life. Robotics is the

868 science of designing, manufacturing, operating, and maintaining of robots and other computer

869 activities to imitate human behavior – very typical in reinforcement machine learning problems

870 [149]. The limited application in robotic construction [21] especially in GB presents

871 opportunities for future research. There are opportunities in using smart robotics – either

872 ground or aerial robots – to rapidly drive semi- or fully-autonomous construction applications

873 such as repetitive and very dangerous construction tasks [20,21]. Hence, it will beneficial to

874 develop cobots (collaborative robots) to work with humans to deliver the sustainable objectives

875 of GB projects such as resource optimization, improved productivity, efficiency, safety and

876 quality [21]. Besides, aerial robots – unmanned aerial vehicles (UAV) with integrated image

877 acquisition systems (i.e., camera, laser scanner, go-pros) – can be trained with machine learning

878 for green retrofits, site monitoring, and structural health monitoring, since they can ensure an

879 economic, simpler, and effective process [20]. For instance, UAVs can fly into an existing

880 building to capture real-time high-definition videos and images, and remotely carry-out laser

881 scanning [20] necessary for green retrofits. Moreover, smart robotics can be integrated with the

882 emerging 4D printing technology [20] for large scale additive manufacturing of green

883 construction elements [159] and even modular integrated construction [21]. 4D printing are 3D

884 printed objects capable of evolving their behaviour and shape within a period to respond to

885 external stimuli including light, heat, temperature, etc. [20]. Hence, future studies that explore

886 the integration of smart robotics and 4D printing would further our knowledge on GB

47
887 automation systems and intelligent optimization. Moreover, integrating BIM with smart

888 robotics and 4D printing could improve the level of automation, efficiency, productivity,

889 safety, and quality of the AEC sector.

890 6.4. Legal, ethical, and moral responsibilities of AI-in-GB

891 From the review, it is observed that AI offers the GB industry a lot of advantages and

892 opportunities for digitization, automation, augmentation, and optimization. GBs today provide

893 a large amount of accumulated valuable data which undergoes continual assessment and

894 analysis to assist facility managers in improving operational performance and reducing

895 building energy waste [81]. As such, modern technology has generated a phenomenal wealth

896 of data and the means of storing it in a quick and easily retrievable manner [151]. Given the

897 growth in data collection and the application of AI, there is the need to investigate further the

898 legal, ethical, social, and moral considerations for AI-in-GB to inform stakeholders such as

899 project developers, software developers, financial institutions, project owners, GB occupants

900 or tenants. Such awareness is necessary in making informed AI-in-GB decisions [160] without

901 compromising legal, ethical, and moral duties. At present, the literature has focused on the

902 technological development and application of AI-in-GB, with little to no attention on the legal,

903 ethical, and moral risks associated with the technological advancement. Therefore, urgent

904 attention is needed to inform regulators and stakeholders of present regulations and its

905 challenges for the industry. AI-in-GB related legal issues such as risk allocation, insurance

906 coverage, civil wrongs (tort, and breach of contract, warranty, and trust) present opportunities

907 for future research. Additionally, future research could consider how the AI is designed,

908 operates, and learns and how the algorithms work in the context of ethical and social problems.

909 Likewise, due to AI-internet integration, and its associated challenges such as data

910 manipulation, exploitation by hackers, cybercrimes, and privacy intrusion [149], further study

911 on AI-in-GB data protection is needed. Moreover, since the government plays critical role in

48
912 promoting GB, further research is required to shape the role of government in addressing the

913 ethical and legal challenges, particularly around the responsibility for and explainability of

914 decisions made by an automation AI system [150]. Further understanding on how government

915 can develop adequate policies, regulations, ethical guidance, and legal framework to prevent

916 misuses of AI-in-GB and their potential disastrous consequences on both individual and

917 societal levels is required. Other critical issues worth exploring include: (1) accountability, (2)

918 laws needed in an autonomous decision-making environment (3) will AI technology replace or

919 assist the decision maker? (4) ultimate responsibility of the designer (will they be responsible

920 only for providing data to feed the algorithm?) and (5) does the algorithm know what is best

921 for us? Future studies are needed to explore these critical legal, ethical, moral, and social issues

922 related to AI-in-GB.

923 7. Conclusions

924 This study reviewed the existing literature on AI-in-GB for the first time to identify research

925 trends and knowledge gaps that may be addressed in future studies. Theoretically, this study

926 presents the key research hotspots, strengths and limitations of AI-in-GB, and gaps in existing

927 studies to inform the directions for future research efforts. Unlike previous isolated systematic

928 or bibliometric reviews, this study leverages the benefits of a mixed-method bibliometric-

929 systematic review to overcome the weakness of review methods when used in isolation. This

930 was effective in limiting subjectivity in the analysis and the ability to replicate similar studies

931 in future. It is observed that early research focused on expert knowledge discovery and fuzzy

932 rules. With the advancement in AI technology coupled with the now-available high volume

933 and rate of generating GB big data, data mining algorithms have become handy in discovering

934 hidden knowledge from such large datasets for future GB decisions. Moreover, intelligent

935 optimization has significantly transformed the ability to locate pareto optimal solutions for GB

936 using AI techniques over traditional approaches. Knowledge from both data mining and

49
937 intelligent optimization of GBs provide a vast quantity of data on existing building operations,

938 allowing GBs to be automatically and intelligently monitored and controlled in real-time. The

939 findings of the study set the tone for further studies by providing paths and recommendations

940 for future studies in AI-in-GB. Practically, this research provides an up-to-date reference for

941 AI-in-GB. This study identified emerging AI technologies, techniques and algorithms

942 introduced in GB-research. As a result, the findings may be used as a reference for practitioners

943 and policymakers to assess their level of development and readiness to embrace AI-in-GB

944 techniques and practices. The strengths and limitations of AI-in-GB identified is this study

945 would be a useful reference for practitioners to ascertain the potential opportunities and risks

946 in AI-in-GB. For construction stakeholders, and first-time developers of GBs, the adoption of

947 AI-in-GB would be a great addition towards efficiency and sustainability. It is suggested that

948 future research would explore the integration of AI and emerging technologies such as digital

949 twins and AIoT, blockchain, and robotics and 4D printing. Nonetheless, due to the

950 digitalization and automation opportunities present in AI-in-GB, there is urgent need to

951 investigate the legal, ethical, and moral challenges associated of AI technologies.

952 Despite the contributions, this study still has limitations which should be taken into

953 consideration when interpreting the study findings. The analysis was based on a single database

954 (Scopus) which may affect the coverage of publications in the study area. Moreover, the

955 research was restricted to journal articles. To improve this study, future studies may combine

956 different databases and document types. Additionally, the literature search was conducted using

957 certain keywords, which might not reflect the complete picture of the research areas. Future

958 studies may include more keywords.

959 Acknowledgements

960 The authors thank the Department of Building and Real Estate of The Hong

961 Kong Polytechnic University for funding this research. We are extremely grateful to the

50
962 Editors and Reviewers whose constructive and invaluable comments and suggestions helped

963 improve the quality of this paper significantly.

964 Appendix A. List of papers included in the qualitative-systematic analysis.

S/N Authors Title Citations Country a
Top 20 most-cited AI-in-GB publications b
1 Wang et al. [41] Applying multi-objective GA in GB design optimization 478 Canada
2 Juan et al. [126] A hybrid decision support system for sustainable office building 251 Taiwan
renovation and energy performance improvement
3 Yu et al. [46] Application of multi-objective GA to optimize energy efficiency and 197 China
thermal comfort in building design
4 Zhao et al. [99] Occupant behaviour and schedule modeling for building energy 186 US
simulation through office appliance power and consumption data mining
5 Akadiri et al. [43] Multi-criteria evaluation model for the selection of sustainable materials 175 UK
for building projects
6 Wang et al. [40] Floor shape optimization for GB design 170 Canada
7 Panda et al. [159] c Additive manufacturing of geopolymer for sustainable built environment 167 Singapore
8 Méndez et al. [113] The early design stage of a building envelope: multi-objective search 166 Italy
through heating, cooling, and lighting energy performance analysis
9 Zhao et al. [45] A fuzzy synthetic evaluation approach for risk assessment: a case of 127 Singapore
Singapore's green projects
10 Tam et al. [71] Green construction assessment for environmental management in the 118 Hong Kong
construction industry of Hong Kong
11 Marinakis et al. An integrated system for buildings' energy-efficient automation: 112 Greece
[161] c Application in the tertiary sector
12 Asl et al. [125] BPot: A framework for BIM-based performance optimization 103 US
13 Chen et al. [48] Optimal control of HVAC and window systems for natural ventilation 102 US
through reinforcement learning
14 Wu and Issa [162] BIM execution planning in GB projects: LEED as a use case 94 US
15 Yezioro et al. [87] An applied artificial intelligence approach towards assessing building 90 US
performance simulation tools
16 Khoshnava et al. [68] Rank of GB material criteria based on the three pillars of sustainability 85 Malaysia
using the hybrid MCDM method
17 Liu et al. [111] Building information modeling-based building design optimization for 83 China
sustainability
18 Yan et al. [134] ARX model-based fault detection and diagnosis for chillers using SVMs 83 United Arab
Emirates
19 Tatari and Kucukvar Cost premium prediction of certified GBs: a neural network approach 82 US
[84]
20 Yang et al. [136] Reinforcement learning for optimal control of low exergy buildings 82 Switzerland
Recent five publication from the top five AI-in-GB outlets
Building and Environment
21 Figueiredo et al. [64] Sustainable material choice for construction projects: A Life Cycle 9 Brazil
Sustainability Assessment framework based on BIM and Fuzzy-AHP
22 Wang et al. [88] The influencing factors of China's GB development: An analysis using 5 China
RBF-WINGS method
23 Park and Park [135] Comparative analysis on predictability of natural ventilation rate based 5 South Korea
on machine learning algorithms
24 Wang et al. [116] Multi-objective optimization (MOO) for high-rise residential buildings’ 1 China
layout centered on daylight, visual, and outdoor thermal metrics in
China
25 Guo et al. [104] Occupants’ satisfaction with LEED- and non-LEED-certified apartments 0 US
using social media data
Energy and Buildings
26 Lin et al. [123] Optimal design of a thermal energy storage system using phase change 16 Australia
materials for a net-zero energy Solar Decathlon house
27 Gonçalves et al. One step forward toward smart city Utopia: Smart building energy 9 Portugal
[137] management based on adaptive surrogate modeling
28 Lu et al. [44] Multidimensional performance-based evaluation method of high- 0 China
performance cold source in GB
29 Wang et al. [118] Experimental study and multi-objective optimisation of a novel integral 0 China
thermoelectric wall
30 Briga-Sá et al. [79] Trombe wall thermal performance: Data mining techniques for indoor 0 Portugal
temperatures and heat flux forecasting

51
Journal of Cleaner Production
31 Shahmansouri et al. ANN model to predict the compressive strength of eco-friendly 35 Iran
[93] geopolymer concrete incorporating silica fume and natural zeolite
32 Ganesh and Development of high performance sustainable optimized fiber 18 India
Muthukannan [92] reinforced geopolymer concrete and prediction of compressive strength
33 Yadegaridehkordi et Assessment of sustainability indicators for GB manufacturing using 14 Malaysia
al. [69] fuzzy MCDM approach
34 Mohandes and Developing a Holistic Occupational Health and Safety risk assessment 6 Hong Kong
Zhang [67] model: An application to a case of sustainable construction project
35 Negash et al. [70] Sustainable construction and demolition waste management in 4 Taiwan
Somaliland: Regulatory barriers lead to technical and environmental
barriers
Applied Energy
36 Fan and Xia [127] Building retrofit optimization models using notch test data considering 16 China
energy performance certificate compliance
37 Westermann et al. Using a deep temporal convolutional network as a building energy 10 Canada
[138] surrogate model that spans multiple climate zones
38 Ding et al. [139] Evolutionary double attention-based long short-term memory model for 5 China
building energy prediction: Case study of a GB
39 Naji et al. [129] Multi-objective optimisations of envelope components for a 2 Australia
prefabricated house in six climate zones
Automation in Construction
40 Fernandez-Ceniceros Decision support model for one-way floor slab design: A sustainable 21 Spain
et al. [90] approach
41 Karatas and El- Optimizing trade-offs among housing sustainability objectives 17 US
Rayes [115]
42 Hong et al. [103] Automated management of GB material information using web crawling 10 South Korea
and ontology
43 Martínez-Rocamora Environmental benchmarking of building typologies through BIM-based 0 Spain
et al. [101] combinatorial case studies
Manual backward and forward snowball citation analysis
44 Platt et al. [122] Adaptive HVAC zone modeling for sustainable buildings 69 Australia
45 Inyim et al. [117] Integration of building information modeling and economic and 66 US
environmental impact analysis to support sustainable building design
46 Marzouk et al. [121] BIM-based approach for optimizing life cycle costs of sustainable 64 Saudi
buildings Arabia
47 Nilashi et al. [58] A knowledge-based expert system for assessing the performance level of 57 Malaysia
GBs
48 Cheng and Ma [97] A non-linear CBR approach for retrieval of similar cases and selection 55 Hong Kong
of target credits in LEED projects
49 Kasinalis et al. [120] Framework for assessing the performance potential of seasonally 53 Netherlands
adaptable facades using multi-objective optimization
50 Ma and Cheng [106] Data-driven study on the achievement of LEED credits using percentage 50 Hong Kong
of average score and association rule analysis
51 Chen and Yang [98] A multi-stage optimization of passively designed high-rise residential 50 Hong Kong
buildings in multiple building operation scenarios
52 Ren and Cao [89] Implementation and visualization of artificial intelligent ventilation 49 China
control system using fast prediction models and limited monitoring data
53 Seo et al. [60] Fuzzy decision-making tool for environmental sustainable buildings 46 Japan
54 Jun and Cheng [100] Selection of target LEED credits based on project information and 42 Hong Kong
climatic factors using data mining techniques
55 Chen et al. [124] A holistic passive design approach to optimize indoor environmental 41 Hong Kong
quality of a typical residential building
56 Allen et al. [132] Fuzzy neural network-based health monitoring for HVAC system 39 US
variable-air-volume unit
57 Son and Kim [85] Early prediction of the performance of GB projects using pre-project 33 South Korea
planning variables: data mining approaches
58 Cheng and Ma [80] A data-driven study of important climate factors on the achievement of 33 Hong Kong
LEED-EB credits
59 Ma and Cheng [102] Identification of the numerical patterns behind the leading counties in 32 Hong Kong
the US local GB markets using data mining
60 Liu and Hu [105] Attention and sentiment of Chinese public toward GBs based on Sina 32 China
Weibo
61 Vakili-Ardebili and Application of fuzzy techniques to develop an assessment framework for 28 UK
Boussabaine [61] building design eco-drivers
62 Yin and Li [65] Matching management of supply and demand of GB technologies based 26 China
on a novel matching method with intuitionistic fuzzy sets

52
 63 Jalaei et al. [96] An integrated BIM-LEED application to automate sustainable design 21 Canada
assessment framework at the conceptual stage of building projects
64 Zhao et al. [107] CBR approach for supporting building green retrofit decisions 21 China
65 Atis and Ekren [133] Development of an outdoor lighting control system using expert system 20 Turkey
66 Juan et al. [86] Identifying customer behavioral factors and price premiums of GB 18 Taiwan
purchasing
67 Yin and Li [66] Academic research institutes-construction enterprises linkages for the 16 China
development of urban GB: Selecting management of GB technologies
innovation partner
68 Zhu et al. [131] Data-driven building load profiling and energy management 13 China
69 Wang et al. [62] Green performance evaluation system for energy-efficiency-based 13 Canada
planning for construction site layout
70 Wang and Wei [112] Design optimization of office building envelope based on QGA for 13 China
energy conservation
71 Lin et al. [114] Multi-objective optimization design of GB envelopes and air 13 Taiwan
conditioning systems for energy conservation and CO2 emission
reduction
72 Bhatt and Macwan Fuzzy Logic and Analytic Hierarchy Process-Based Conceptual Model 10 India
[63] for Sustainable Commercial Building Assessment for India
73 Abediniangerabi et A data-driven framework for energy-conscious design of building 9 US
al. [83] facade systems
74 May Tzuc et al. [91] Modeling of hygrothermal behavior for green facade's concrete wall 5 Finland
exposed to Nordic climate using AI and global sensitivity analysis
75 Wen et al. [95] Assessing the Effectiveness of BIM in Developing GBs from a 4 China
Lifecycle Perspective
76 Lee and Lee [119] Optimization of apartment-complex layout planning for daylight 2 South Korea
accessibility in a high-density city with a temperate climate
77 Elshaboury and Optimizing construction and demolition waste transportation for 1 Egypt
Marzouk [128] sustainable construction
78 Bajno et al. [94] Old and modern wooden buildings in the context of sustainable 0 Poland
development
965 a Where authors are affiliated to different countries, the country of the first author is listed.

966 b The list includes only reviewed articles in systematic analysis.

967 c These two articles were among the top 20 most cited AI-in-GB studies but were not included in systematic analysis.

968
969 References
970 [1] UNEP, 2021 Global Status Report for Buildings and Construction, UNEP - United
971 Nations Environment Programme. (2021). http://www.unep.org/resources/report/2021-
972 global-status-report-buildings-and-construction (accessed October 25, 2021).
973 [2] US Environmental Protection Agency, Definition of Green Building., (2016).
974 https://archive.epa.gov/greenbuilding/web/html/about.html (accessed April 4, 2021).
975 [3] WorldGBC, About Green Buildings, (2021). https://www.worldgbc.org/about-green-
976 building (accessed February 24, 2021).
977 [4] A. Darko, A.P.C. Chan, D.-G. Owusu-Manu, E.E. Ameyaw, Drivers for implementing
978 green building technologies: An international survey of experts, Journal of Cleaner
979 Production. 145 (2017), pp. 386–394. https://doi.org/10.1016/j.jclepro.2017.01.043.
980 [5] A.P.C. Chan, A. Darko, A.O. Olanipekun, E.E. Ameyaw, Critical barriers to green
981 building technologies adoption in developing countries: The case of Ghana, Journal of
982 Cleaner Production. 172 (2018), pp. 1067–1079.
983 https://doi.org/10.1016/j.jclepro.2017.10.235.
984 [6] J. Zuo, Z.-Y. Zhao, Green building research–current status and future agenda: A review,
985 Renewable Sustainable Energy Reviews. 30 (2014), pp. 271–281.
986 https://doi.org/10.1016/j.rser.2013.10.021.
987 [7] A. Darko, A.P.C. Chan, Critical analysis of green building research trend in construction
988 journals, Habitat International. 57 (2016), pp. 53–63.
989 https://doi.org/10.1016/j.habitatint.2016.07.001.
990 [8] C. Debrah, A.P.C. Chan, A. Darko, Green finance gap in green buildings: A scoping
991 review and future research needs, Building and Environment. 207 (2022), pp. 108443.
992 https://doi.org/10.1016/j.buildenv.2021.108443.
53
 993 [9] D. Yu, H. Liao, Visualization and quantitative research on intuitionistic fuzzy studies,
994 Journal of Intelligent & Fuzzy System. 30 (2016), pp. 3653–3663.
995 https://doi.org/10.3233/IFS-162111.
996 [10] A. Darko, A.P.C. Chan, X. Huo, D.-G. Owusu-Manu, A scientometric analysis and
997 visualization of global green building research, Building and Environment. 149 (2019),
998 pp. 501–511. https://doi.org/10.1016/j.buildenv.2018.12.059.
999 [11] X. Zhao, J. Zuo, G. Wu, C. Huang, A bibliometric review of green building research
1000 2000–2016, Architectural Science Review. 62 (2019), pp. 74–88.
1001 https://doi.org/10.1080/00038628.2018.1485548.
1002 [12] J. McCarthy, The inversion of functions defined by Turing machines, In Automa Studies
1003 (AM-34). 34 (1956), pp. 177–181. Princeton University Press
1004 [13] P. Whig, Artificial Intelligence and Machine Learning in Business, International Journal
1005 of Integrated Education. 2 (2019), pp. 334128.
1006 http://journals.researchparks.org/index.php/IJIE/article/view/516
1007 [14] A. Jeerage, B. Erwine, S. Mallory, V. Agarwal, Predicting green building performance
1008 over time: data mining untapped information in LEED, WIT Transactions on Ecology
1009 and the Environment. 128 (2010), pp. 355–365. https://doi.org/10.2495/ARC100301
1010 [15] H. Chen, B. Zhou, Environmental performance evaluation of green buildings based on
1011 machine learning and IOT systems, Microprocessors and Microsystems. (2020), pp.
1012 103496. http://doi.org/10.1016/j.micpro.2020.103496
1013 [16] Y. Lu, Z. Wu, R. Chang, Y. Li, Building Information Modeling (BIM) for green
1014 buildings: A critical review and future direction. Automation in Construction, (2017),
1015 pp. 134-138. https://doi.org/10.1016/j.autcon.2017.08.024
1016 [17] M. Bilal, L.O. Oyedele, J. Qadir, K. Munir, S.O. Ajayi, O.O. Akinade, H.A. Owolabi,
1017 H.A. Alaka, M. Pasha, Big Data in the construction industry: A review of present status,
1018 opportunities, and future trends, Advances Engineering Informatics. 30 (2016), pp. 500–
1019 521. https://doi.org/10.1016/j.aei.2016.07.001
1020 [18] V. Faghihi, A. Nejat, K.F. Reinschmidt, J.H. Kang, Automation in construction
1021 scheduling: a review of the literature, The International Journal of Advanced
1022 Manufacturing Technology. 81 (2015), pp. 1845–1856. https://doi.org/10.1007/s00170-
1023 015-7339-0
1024 [19] Z. Irani, M.M. Kamal, Intelligent Systems Research in the Construction Industry, Expert
1025 Systems with Applications. 41 (2014), pp. 934–950.
1026 https://doi.org/10.1016/j.eswa.2013.06.061.
1027 [20] Y. Pan, L. Zhang, Roles of artificial intelligence in construction engineering and
1028 management: A critical review and future trends, Automation in Construction. 122
1029 (2021), pp. 103517. https://doi.org/10.1016/j.autcon.2020.103517.
1030 [21] A. Darko, A.P.C. Chan, M.A. Adabre, D.J. Edwards, M.R. Hosseini, E.E. Ameyaw,
1031 Artificial intelligence in the AEC industry: Scientometric analysis and visualization of
1032 research activities, Automation in Construction. 112 (2020), pp. 103081.
1033 https://doi.org/10.1016/j.autcon.2020.103081.
1034 [22] M. Heyvaert, K. Hannes, P. Onghena, Using Mixed Methods Research Synthesis for
1035 Literature Reviews: The Mixed Methods Research Synthesis Approach, SAGE
1036 Publications, (2016). ISBN 978143358291
1037 [23] A. Harden, J. Thomas, Mixed Methods and Systematic Reviews: Examples and
1038 Emerging Issues, in: SAGE Handbook of Mixed Methods in Social & Behavioral
1039 Research, SAGE Publications, Incorporation, 2455 Teller Road, Thousand
1040 Oaks California 91320 United States, (2010): pp. 749–774.
1041 https://doi.org/10.4135/9781506335193.n29.

54
1042 [24] P. Pluye, Q.N. Hong, Combining the Power of Stories and the Power of Numbers: Mixed
1043 Methods Research and Mixed Studies Reviews, Annual Review of Public Health. 35
1044 (2014) pp. 29–45. https://doi.org/10.1146/annurev-publhealth-032013-182440.
1045 [25] X. Yin, H. Liu, Y. Chen, M. Al-Hussein, Building information modelling for off-site
1046 construction: Review and future directions, Automation in Construction. 101 (2019), pp.
1047 72–91. https://doi.org/10.1016/j.autcon.2019.01.010.
1048 [26] N.J. van Eck, L. Waltman, Visualizing Bibliometric Networks, in: Y. Ding, R. Rousseau,
1049 D. Wolfram (Eds.), Measuring Scholarly Impact: Methods and Practice, Springer
1050 International Publishing, Cham, (2014): pp. 285–320. https://doi.org/10.1007/978-3-
1051 319-10377-8_13.
1052 [27] A. Darko, A.P.C. Chan, Applying science mapping in built environment research,
1053 Routledge, (2021). https://doi.org/10.1201/9781003000532-9.
1054 [28] M.J. Cobo, A.G. López‐Herrera, E. Herrera‐Viedma, F. Herrera, Science mapping
1055 software tools: Review, analysis, and cooperative study among tools, Journal of the
1056 American Society for Information Science and Technology. 62 (2011), pp. 1382–1402.
1057 https://doi.org/10.1002/asi.21525.
1058 [29] M. Oraee, M.R. Hosseini, E. Papadonikolaki, R. Palliyaguru, M. Arashpour,
1059 Collaboration in BIM-based construction networks: A bibliometric-qualitative literature
1060 review, International Journal of Project Management. 35 (2017), pp. 1288–1301.
1061 https://doi.org/10.1016/j.ijproman.2017.07.001.
1062 [30] J.C. Greene, V.J. Caracelli, W.F. Graham, Toward a Conceptual Framework for Mixed-
1063 Method Evaluation Designs, Educational Evaluation and Policy Analysis. 11 (1989), pp.
1064 255–274. https://doi.org/10.3102/01623737011003255.
1065 [31] L.I. Meho, Y. Rogers, Citation counting, citation ranking, and h-index of human-
1066 computer interaction researchers: A comparison of Scopus and Web of Science, Journal
1067 of American Society for Information Science and Technology. 59 (2008), pp. 1711–
1068 1726. https://doi.org/10.1002/asi.20874.
1069 [32] C. Collins, D. Dennehy, K. Conboy, P. Mikalef, Artificial intelligence in information
1070 systems research: A systematic literature review and research agenda, I International
1071 Journal of Information Management. 60 (2021), pp. 102383.
1072 https://doi.org/10.1016/j.ijinfomgt.2021.102383.
1073 [33] M.R. Hosseini, I. Martek, E.K. Zavadskas, A.A. Aibinu, M. Arashpour, N. Chileshe,
1074 Critical evaluation of off-site construction research: A Scientometric analysis,
1075 Automation in Construction. 87 (2018), pp. 235–247.
1076 https://doi.org/10.1016/j.autcon.2017.12.002.
1077 [34] C. Chen, CiteSpace II: Detecting and visualizing emerging trends and transient patterns
1078 in scientific literature, International Journal of American Society for Information
1079 Science and Technology. 57 (2006), pp. 359–377. https://doi.org/10.1002/asi.20317.
1080 [35] M. Bastian, S. Heymann, M. Jacomy, Gephi: An Open Source Software for Exploring
1081 and Manipulating Networks, Proceedings of the Third AAAI Conference on Weblogs
1082 and Social Media. 3 (2009), pp. 361-362.
1083 https://ojs.aaai.org/index.php/ICWSM/article/view/13937 (accessed March 12, 2021).
1084 [36] H. Li, Q. Shen, Supporting the decision-making process for sustainable housing,
1085 Construction Management and Economics. 20 (2002), pp. 387–390.
1086 https://doi.org/10.1080/01446190210146812
1087 [37] J.F. Hair, Multivariate data analysis: A Global Perspective. 7th ed. Upper Saddle River:
1088 Prentice Hall, (2009). ISBN-13: 978-0138132637
1089 [38] A.A.T. Fernandes, D.B. Figueiredo, E.C. da Rocha, W. da S. Nascimento, Read this
1090 paper if you want to learn logistic regression, Revista de Sociologia. e Política. 28
1091 (2021), pp. 1-19. https://doi.org/10.1590/1678-987320287406en

55
1092 [39] N.J. van Eck, L. Waltman, How to normalize cooccurrence data? An analysis of some
1093 well-known similarity measures, Journal of American Society for Information Science
1094 and Technology. 60 (2009), pp. 1635–1651. https://doi.org/10.1002/asi.21075.
1095 [40] W. Wang, H. Rivard, R. Zmeureanu, Floor shape optimization for green building design,
1096 Advances Engineering Informatics. 20 (2006), pp. 363–378.
1097 https://doi.org/10.1016/j.aei.2006.07.001.
1098 [41] W. Wang, R. Zmeureanu, H. Rivard, Applying multi-objective genetic algorithms in
1099 green building design optimization, Building and Environment. 40 (2005), pp. 1512–
1100 1525. https://doi.org/10.1016/j.buildenv.2004.11.017.
1101 [42] L.A. Zadeh, Information and control, Fuzzy Sets. In Fuzzy sets, fuzzy logic, and fuzzy
1102 systems. 8 (1965), pp. 338–353. https://doi.org/10.1142/9789814261302_0021
1103 [43] P.O. Akadiri, P.O. Olomolaiye, E.A. Chinyio, Multi-criteria evaluation model for the
1104 selection of sustainable materials for building projects, Automation in Construction. 30
1105 (2013), pp. 113–125. https://doi.org/10.1016/j.autcon.2012.10.004
1106 [44] S. Lu, Y. Liu, Y. Li, R. Wang, Multidimensional performance-based evaluation method
1107 of high-performance cold source in green building, Energy and Buildings. 231 (2021),
1108 pp. 110618. https://doi.org/10.1016/j.enbuild.2020.110618
1109 [45] X. Zhao, B.-G. Hwang, Y. Gao, A fuzzy synthetic evaluation approach for risk
1110 assessment: a case of Singapore’s green projects, Journal of Cleaner Production. 115
1111 (2016), pp. 203–213. https://doi.org/10.1016/j.jclepro.2015.11.042.
1112 [46] W. Yu, B. Li, H. Jia, M. Zhang, D. Wang, Application of multi-objective genetic
1113 algorithm to optimize energy efficiency and thermal comfort in building design, Energy
1114 and Buildings. 88 (2015), pp. 135–143. https://doi.org/10.1016/j.enbuild.2014.11.063.
1115 [47] R.J. Cole, Z. Brown, Reconciling human and automated intelligence in the provision of
1116 occupant comfort, Intelligent Buildings International. 1 (2009), pp. 39–55.
1117 https://doi.org/10.3763/inbi.2009.0007
1118 [48] Y. Chen, L.K. Norford, H.W. Samuelson, A. Malkawi, Optimal control of HVAC and
1119 window systems for natural ventilation through reinforcement learning, Energy and
1120 Buildings. 169 (2018), pp. 195–205. https://doi.org/10.1016/j.enbuild.2018.03.051
1121 [49] C. Chen, CiteSpace Manual, (2014). College of Computing and Informatics.
1122 http://cluster.ischool.drexel.edu/%7Ecchen/citespace/CiteSpaceManual.pdf.
1123 [50] C. Prell, Social Network Analysis: History, Theory and Methodology, SAGE,
1124 Publications, (2012). ISBN 978-1-4129-4715-2
1125 [51] J. Vincent, Artificial intelligence research continues to grow as China overtakes US in
1126 AI journal citations, The Verge. (2021).
1127 https://www.theverge.com/2021/3/3/22310840/ai-research-global-growth-china-us-
1128 paper-citations-index-report-2020 (accessed December 8, 2021).
1129 [52] B. Crew, Google Scholar reveals its most influential papers for 2020, Nature Index.
1130 (2020). https://www.natureindex.com/news-blog/google-scholar-reveals-most-
1131 influential-papers-research-citations-twenty-twenty (accessed December 7, 2021).
1132 [53] S.A. Aaronson, America’s Uneven Approach to AI and Its Consequence, Great
1133 Decisions. (2020), pp. 87-96. Available at SSRN: https://ssrn.com/abstract=3589847
1134 [54] Y. Ding, Scientific collaboration and endorsement: Network analysis of coauthorship
1135 and citation networks, Journal of Informetrics. 5 (2011), pp. 187–203.
1136 https://doi.org/10.1016/j.joi.2010.10.008.
1137 [55] J.W. Kirchherr, R. van Santen, Research on the Circular Economy: A Critique of the
1138 Field, Resources, Conservation and Recycling. 151 (2019).
1139 https://doi.org/10.1016/j.resconrec.2019.104480.

56
1140 [56] P. Antwi-Afari, S.T. Ng, Md.U. Hossain, A review of the circularity gap in the
1141 construction industry through scientometric analysis, Journal of Cleaner Production. 298
1142 (2021), pp. 126870. https://doi.org/10.1016/j.jclepro.2021.126870.
1143 [57] C. Wohlin, Guidelines for snowballing in systematic literature studies and a replication
1144 in software engineering, In Proceedings of the 18th International Conference on
1145 Evaluation and Assessment in Software Engineering, Association for Computing
1146 Machinery, New York, NY, USA, (2014), pp. 1–10.
1147 https://doi.org/10.1145/2601248.2601268.
1148 [58] M. Nilashi, R. Zakaria, O. Ibrahim, M. Z. Abd. Majid, R. Mohamad Zin, M.W. Chugtai,
1149 N.I. Zainal Abidin, S.R. Sahamir, D. Aminu Yakubu, A knowledge-based expert system
1150 for assessing the performance level of green buildings, Knowledge-Based Systems. 86
1151 (2015), pp. 194–209. https://doi.org/10.1016/j.knosys.2015.06.009.
1152 [59] S. Mu, H. Cheng, M. Chohr, W. Peng, Assessing risk management capability of
1153 contractors in subway projects in mainland China, International Journal of Project
1154 Management. 32 (2014), pp. 452–460. https://doi.org/10.1016/j.ijproman.2013.08.007.
1155 [60] S. Seo, T. Aramaki, Y. Hwang, K. Hanaki, Fuzzy Decision-Making Tool for
1156 Environmental Sustainable Buildings, Journal of Construction and Engineering
1157 Management. 130 (2004), pp. 415–423. https://doi.org/10.1061/(ASCE)0733-
1158 9364(2004)130:3(415).
1159 [61] A. Vakili-Ardebili, A.H. Boussabaine, Application of fuzzy techniques to develop an
1160 assessment framework for building design eco-drivers, Building and Environment. 42
1161 (2007), pp. 3785–3800. https://doi.org/10.1016/j.buildenv.2006.11.017.
1162 [62] C.C. Wang, S.M. Sepasgozar, M. Wang, J. Sun, X. Ning, Green performance evaluation
1163 system for energy-efficiency-based planning for construction site layout, Energies. 12
1164 (2019), pp. 4620. https://doi.org/10.3390/en12244620
1165 [63] R. Bhatt, J.E.M. Macwan, Fuzzy logic and analytic hierarchy process–based conceptual
1166 model for sustainable commercial building assessment for India, Journal of Architectural
1167 Engineering. 22 (2016), pp. 04015009. https://doi.org/10.1061/(ASCE)AE.1943-
1168 5568.0000184
1169 [64] K. Figueiredo, R. Pierott, A.W. Hammad, A. Haddad, Sustainable material choice for
1170 construction projects: A life cycle sustainability assessment framework based on BIM
1171 and Fuzzy-AHP, Building and Environment. (2021), pp. 107805.
1172 https://doi.org/10.1016/j.buildenv.2021.107805
1173 [65] S. Yin, B. Li, Matching management of supply and demand of green building
1174 technologies based on a novel matching method with intuitionistic fuzzy sets, Journal of
1175 Cleaner Production. 201 (2018), pp. 748–763.
1176 https://doi.org/10.1016/j.jclepro.2018.08.055.
1177 [66] S. Yin, B. Li, Academic research institutes-construction enterprises linkages for the
1178 development of urban green building: Selecting management of green building
1179 technologies innovation partner, Sustainable Cities Society. 48 (2019), pp. 101555.
1180 https://doi.org/10.1016/j.scs.2019.101555
1181 [67] S.R. Mohandes, X. Zhang, Developing a Holistic Occupational Health and Safety risk
1182 assessment model: An application to a case of sustainable construction project, Journal
1183 of Cleaner Production. 291 (2021), pp. 125934.
1184 https://doi.org/10.1016/j.jclepro.2021.125934
1185 [68] S.M. Khoshnava, R. Rostami, A. Valipour, M. Ismail, A.R. Rahmat, Rank of green
1186 building material criteria based on the three pillars of sustainability using the hybrid
1187 multi criteria decision making method, Journal of Cleaner Production. 173 (2018), pp.
1188 82–99. https://doi.org/10.1016/j.jclepro.2016.10.066

57
1189 [69] E. Yadegaridehkordi, M. Hourmand, M. Nilashi, E. Alsolami, S. Samad, M. Mahmoud,
1190 A.A. Alarood, A. Zainol, H.D. Majeed, L. Shuib, Assessment of sustainability indicators
1191 for green building manufacturing using fuzzy multi-criteria decision making approach,
1192 Journal of Cleaner Production. 277 (2020), pp. 122905.
1193 https://doi.org/10.1016/j.jclepro.2020.122905
1194 [70] Y.T. Negash, A.M. Hassan, M.-L. Tseng, K.-J. Wu, M.H. Ali, Sustainable construction
1195 and demolition waste management in Somaliland: Regulatory barriers lead to technical
1196 and environmental barriers, Journal of Cleaner Production. 297 (2021) pp. 126717.
1197 https://doi.org/10.1016/j.jclepro.2021.126717
1198 [71] C.M. Tam, V.W.Y. Tam, W.S. Tsui, Green construction assessment for environmental
1199 management in the construction industry of Hong Kong, International Journal of Project
1200 Management. 22 (2004), pp. 563–571. https://doi.org/10.1016/j.ijproman.2004.03.001.
1201 [72] T.L. Saaty, What is the Analytic Hierarchy Process?, in: G. Mitra, H.J. Greenberg, F.A.
1202 Lootsma, M.J. Rijkaert, H.J. Zimmermann (Eds.), In Mathematical Models for Decision
1203 Support, Springer, Berlin, Heidelberg, (1988): pp. 109–121.
1204 https://doi.org/10.1007/978-3-642-83555-1_5.
1205 [73] T.L. Saaty, Decision making with dependence and feedback: The analytic network
1206 process, RWS publications Pittsburgh, USA, 4922 (1996). ISBN 0-9620317-9-8
1207 [74] A. Gabus, E. Fontela, World problems, an invitation to further thought within the
1208 framework of DEMA℡, Battelle Geneva Res. Cent. Geneva Switz. (1972) 1–8.
1209 [75] C.-L. Hwang, K. Yoon, Methods for multiple attribute decision making, In Multiple
1210 Attribute Decision Making, Lecture Notes in Economics and Mathematical Systems,
1211 Springer, Berlin, Heidelberg (1981): pp. 58–191. https://doi.org/10.1007/978-3-642-
1212 48318-9_3
1213 [76] S. Opricovic, Multicriteria optimization of civil engineering systems, Faculty of Civil
1214 Engineering. University of Belgrade. 2 (1998), pp. 5–21.
1215 [77] B. Wang, J. Song, J. Ren, K. Li, H. Duan, X. Wang, Selecting sustainable energy
1216 conversion technologies for agricultural residues: A fuzzy AHP-VIKOR based
1217 prioritization from life cycle perspective, Resources, Conservation and Recycling. 142
1218 (2019), pp. 78–87. https://doi.org/10.1016/j.resconrec.2018.11.011.
1219 [78] IBM, Big Data Analytics, (2021). https://www.ibm.com/analytics/hadoop/big-data-
1220 analytics (accessed November 19, 2021).
1221 [79] A. Briga-Sá, D. Leitão, J. Boaventura-Cunha, F.F. Martins, Trombe wall thermal
1222 performance: Data mining techniques for indoor temperatures and heat flux forecasting,
1223 Energy and Buildings. 252 (2021), pp. 111407.
1224 https://doi.org/10.1016/j.enbuild.2021.111407
1225 [80] J.C. Cheng, L.J. Ma, A data-driven study of important climate factors on the
1226 achievement of LEED-EB credits, Building and Environment. 90 (2015), pp. 232–244.
1227 https://doi.org/10.1016/j.buildenv.2014.11.029
1228 [81] P. Palensky, D. Dietrich, Demand side management: Demand response, intelligent
1229 energy systems, and smart loads, IEEE Transactions on Industrial Informatics. 7 (2011),
1230 pp. 381–388. https://doi.org/10.1109/TII.2011.2158841.
1231 [82] A. Kusiak, Smart manufacturing must embrace big data, Nature. 544 (2017), pp. 23–25.
1232 https://doi.org/10.1038/544023a
1233 [83] B. Abediniangerabi, S.M. Shahandashti, A. Makhmalbaf, A data-driven framework for
1234 energy-conscious design of building facade systems, Journal of Building Engineering.
1235 29 (2020), pp. 101172. https://doi.org/10.1016/j.jobe.2020.101172.
1236 [84] O. Tatari, M. Kucukvar, Cost premium prediction of certified green buildings: A neural
1237 network approach, Building and Environment. 46 (2011), 1081–1086.
1238 https://doi.org/10.1016/j.buildenv.2010.11.009.

58
1239 [85] H. Son, C. Kim, Early prediction of the performance of green building projects using
1240 pre-project planning variables: data mining approaches, Journal of Cleaner Production.
1241 109 (2015), pp. 144–151. https://doi.org/10.1016/j.jclepro.2014.08.071.
1242 [86] Y.-K. Juan, Y.-H. Hsu, X. Xie, Identifying customer behavioral factors and price
1243 premiums of green building purchasing, Industrial Marketing Management. 64 (2017),
1244 pp. 36–43. https://doi.org/10.1016/j.indmarman.2017.03.004.
1245 [87] A. Yezioro, B. Dong, F. Leite, An applied artificial intelligence approach towards
1246 assessing building performance simulation tools, Energy and Buildings. 40 (2008), pp.
1247 612–620. https://doi.org/10.1016/j.enbuild.2007.04.014
1248 [88] W. Wang, Z. Tian, W. Xi, Y.R. Tan, Y. Deng, The influencing factors of China’s green
1249 building development: An analysis using RBF-WINGS method, Building and
1250 Environment. 188 (2021), pp. 107425. https://doi.org/10.1016/j.buildenv.2020.107425.
1251 [89] C. Ren, S.-J. Cao, Implementation and visualization of artificial intelligent ventilation
1252 control system using fast prediction models and limited monitoring data, Sustainable
1253 Cities and Society. 52 (2020), pp. 101860. https://doi.org/10.1016/j.scs.2019.101860
1254 [90] J. Fernandez-Ceniceros, R. Fernandez-Martinez, E. Fraile-Garcia, F.J. Martinez-de-
1255 Pison, Decision support model for one-way floor slab design: A sustainable approach,
1256 Automation in Construction. 35 (2013), pp. 460–470.
1257 https://doi.org/10.1016/j.autcon.2013.06.002
1258 [91] O. May Tzuc, O. Rodríguez Gamboa, R. Aguilar Rosel, M. Che Poot, H. Edelman, M.
1259 Jiménez Torres, A. Bassam, Modeling of hygrothermal behavior for green facade’s
1260 concrete wall exposed to nordic climate using artificial intelligence and global sensitivity
1261 analysis, Journal of Building Engineering. 33 (2021), pp. 101625.
1262 https://doi.org/10.1016/j.jobe.2020.101625.
1263 [92] A.C. Ganesh, M. Muthukannan, Development of high performance sustainable
1264 optimized fiber reinforced geopolymer concrete and prediction of compressive strength,
1265 Journal of Cleaner Production. 282 (2021), pp. 124543.
1266 https://doi.org/10.1016/j.jclepro.2020.124543
1267 [93] A.A. Shahmansouri, M. Yazdani, S. Ghanbari, H. Akbarzadeh Bengar, A. Jafari, H.
1268 Farrokh Ghatte, Artificial neural network model to predict the compressive strength of
1269 eco-friendly geopolymer concrete incorporating silica fume and natural zeolite, Journal
1270 of Cleaner Production. 279 (2021), pp. 123697.
1271 https://doi.org/10.1016/j.jclepro.2020.123697.
1272 [94] D. Bajno, A. Grzybowska, \Lukasz Bednarz, Old and Modern Wooden Buildings in the
1273 Context of Sustainable Development, Energies. 14 (2021), pp. 5975.
1274 https://doi.org/10.3390/en14185975
1275 [95] Q. Wen, Z. Li, Y. Peng, B. Guo, Assessing the Effectiveness of Building Information
1276 Modeling in Developing Green Buildings from a Lifecycle Perspective, Sustainability.
1277 12 (2020), pp. 9988. https://doi.org/10.3390/su12239988.
1278 [96] F. Jalaei, F. Jalaei, S. Mohammadi, An integrated BIM-LEED application to automate
1279 sustainable design assessment framework at the conceptual stage of building projects,
1280 Sustainable Cities and Society. 53 (2020), pp. 101979.
1281 https://doi.org/10.1016/j.scs.2019.101979
1282 [97] J.C. Cheng, L.J. Ma, A non-linear case-based reasoning approach for retrieval of similar
1283 cases and selection of target credits in LEED projects, Building and Environment. 93
1284 (2015), pp. 349–361. https://doi.org/10.1016/j.buildenv.2015.07.019
1285 [98] X. Chen, H. Yang, A multi-stage optimization of passively designed high-rise residential
1286 buildings in multiple building operation scenarios, Applied Energy. 206 (2017), pp. 541–
1287 557. https://doi.org/10.1016/j.apenergy.2017.08.204.

59
1288 [99] J. Zhao, B. Lasternas, K.P. Lam, R. Yun, V. Loftness, Occupant behavior and schedule
1289 modeling for building energy simulation through office appliance power consumption
1290 data mining, Energy and Buildings. 82 (2014), pp. 341–355.
1291 https://doi.org/10.1016/j.enbuild.2014.07.033
1292 [100] M.A. Jun, J.C. Cheng, Selection of target LEED credits based on project information
1293 and climatic factors using data mining techniques, Advances Engineering Informatics.
1294 32 (2017), pp. 224–236. https://doi.org/10.1016/j.aei.2017.03.004
1295 [101] A. Martínez-Rocamora, C. Rivera-Gómez, C. Galán-Marín, M. Marrero, Environmental
1296 benchmarking of building typologies through BIM-based combinatorial case studies,
1297 Automation in Construction. 132 (2021), pp. 103980.
1298 https://doi.org/10.1016/j.autcon.2021.103980
1299 [102] J. Ma, J.C. Cheng, Identification of the numerical patterns behind the leading counties
1300 in the US local green building markets using data mining, Journal of Cleaner Production.
1301 151 (2017), pp. 406–418. https://doi.org/10.1016/j.jclepro.2017.03.083
1302 [103] S.-H. Hong, S.-K. Lee, J.-H. Yu, Automated management of green building material
1303 information using web crawling and ontology, Automation in Construction. 102 (2019),
1304 pp. 230–244. https://doi.org/10.1016/j.autcon.2019.01.015
1305 [104] X. Guo, K. Lee, Z. Wang, S. Liu, Occupants’ satisfaction with LEED-and non-LEED-
1306 certified apartments using social media data, Building and Environment. 206 (2021), pp.
1307 108288. https://doi.org/10.1016/j.buildenv.2021.108288
1308 [105] X. Liu, W. Hu, Attention and sentiment of Chinese public toward green buildings based
1309 on Sina Weibo, Sustainable Cities and Society. 44 (2019), pp. 550–558.
1310 https://doi.org/10.1016/j.scs.2018.10.047
1311 [106] J. Ma, J.C. Cheng, Data-driven study on the achievement of LEED credits using
1312 percentage of average score and association rule analysis, Building and Environment.
1313 98 (2016), pp. 121–132. https://doi.org/10.1016/j.buildenv.2016.01.005
1314 [107] X. Zhao, Y. Tan, L. Shen, G. Zhang, J. Wang, Case-based reasoning approach for
1315 supporting building green retrofit decisions, Building and Environment. 160 (2019), pp.
1316 106210. https://doi.org/10.1016/j.buildenv.2019.106210.
1317 [108] M.W. Gardner, S.R. Dorling, Artificial neural networks (the multilayer perceptron)—a
1318 review of applications in the atmospheric sciences, Atmospheric Environment. 32
1319 (1998), pp. 2627–2636. https://doi.org/10.1016/S1352-2310(97)00447-0
1320 [109] C. Cortes, V. Vapnik, Support-vector networks, Machine Learning. 20 (1995), pp. 273–
1321 297. https://doi.org/10.1007/BF00994018
1322 [110] D. Delen, G. Walker, A. Kadam, Predicting breast cancer survivability: a comparison of
1323 three data mining methods, Artificial Intelligence in Medicine. 34 (2005), pp. 113–127.
1324 https://doi.org/10.1016/j.artmed.2004.07.002.
1325 [111] S. Liu, X. Meng, C. Tam, Building information modeling based building design
1326 optimization for sustainability, Energy and Buildings. 105 (2015), pp. 139–153.
1327 https://doi.org/10.1016/j.enbuild.2015.06.037
1328 [112] Y. Wang, C. Wei, Design optimization of office building envelope based on quantum
1329 genetic algorithm for energy conservation, Journal of Building Engineering. 35 (2021),
1330 pp. 102048. https://doi.org/10.1016/j.jobe.2020.102048.
1331 [113] T.M. Echenagucia, A. Capozzoli, Y. Cascone, M. Sassone, The early design stage of a
1332 building envelope: Multi-objective search through heating, cooling and lighting energy
1333 performance analysis, Applied Energy. 154 (2015), pp. 577–591.
1334 https://doi.org/10.1016/j.apenergy.2015.04.090
1335 [114] Y.-H. Lin, M.-D. Lin, K.-T. Tsai, M.-J. Deng, H. Ishii, Multi-objective optimization
1336 design of green building envelopes and air conditioning systems for energy conservation

60
1337 and CO2 emission reduction, Sustainable Cities and Society. 64 (2021), pp. 102555.
1338 https://doi.org/10.1016/j.scs.2020.102555
1339 [115] A. Karatas, K. El-Rayes, Optimizing tradeoffs among housing sustainability objectives,
1340 Automation in Construction. 53 (2015), pp. 83–94.
1341 https://doi.org/10.1016/j.autcon.2015.02.010
1342 [116] S. Wang, Y.K. Yi, N. Liu, Multi-objective optimization (MOO) for high-rise residential
1343 buildings’ layout centered on daylight, visual, and outdoor thermal metrics in China,
1344 Building and Environment. 205 (2021), pp. 108263.
1345 https://doi.org/10.1016/j.buildenv.2021.108263
1346 [117] P. Inyim, J. Rivera, Y. Zhu, Integration of building information modeling and economic
1347 and environmental impact analysis to support sustainable building design, Journal of
1348 Management in Engineering. 31 (2015), pp. A4014002.
1349 https://doi.org/10.1061/(ASCE)ME.1943-5479.0000308
1350 [118] P. Wang, Z. Liu, D. Chen, W. Li, L. Zhang, Experimental study and multi-objective
1351 optimisation of a novel integral thermoelectric wall, Energy and Buildings. 252 (2021),
1352 pp. 111403. https://doi.org/10.1016/j.enbuild.2021.111403
1353 [119] S. Lee, K.S. Lee, Optimization of Apartment-Complex Layout Planning for Daylight
1354 Accessibility in a High-Density City with a Temperate Climate, Energies. 13 (2020), pp.
1355 4172. https://doi.org/10.3390/en13164172
1356 [120] C. Kasinalis, R.C. Loonen, D. Cóstola, J.L.M. Hensen, Framework for assessing the
1357 performance potential of seasonally adaptable facades using multi-objective
1358 optimization, Energy and Buildings. 79 (2014), pp. 106–113.
1359 https://doi.org/10.1016/j.enbuild.2014.04.045
1360 [121] M. Marzouk, S. Azab, M. Metawie, BIM-based approach for optimizing life cycle costs
1361 of sustainable buildings, Journal of Cleaner Production. 188 (2018), pp. 217–226.
1362 https://doi.org/10.1016/j.jclepro.2018.03.280.
1363 [122] G. Platt, J. Li, R. Li, G. Poulton, G. James, J. Wall, Adaptive HVAC zone modeling for
1364 sustainable buildings, Energy and Buildings. 42 (2010), pp. 412–421.
1365 https://doi.org/10.1016/j.enbuild.2009.10.009.
1366 [123] W. Lin, Z. Ma, C. McDowell, Y. Baghi, B. Banfield, Optimal design of a thermal energy
1367 storage system using phase change materials for a net-zero energy Solar Decathlon
1368 house, Energy and Buildings. 208 (2020), pp. 109626.
1369 https://doi.org/10.1016/j.enbuild.2019.109626
1370 [124] X. Chen, H. Yang, K. Sun, A holistic passive design approach to optimize indoor
1371 environmental quality of a typical residential building in Hong Kong, Energy. 113
1372 (2016), pp. 267–281. https://doi.org/10.1016/j.energy.2016.07.058
1373 [125] M.R. Asl, S. Zarrinmehr, M. Bergin, W. Yan, BPOpt: A framework for BIM-based
1374 performance optimization, Energy and Buildings. 108 (2015), pp. 401–412.
1375 https://doi.org/10.1016/j.enbuild.2015.09.011
1376 [126] Y.-K. Juan, P. Gao, J. Wang, A hybrid decision support system for sustainable office
1377 building renovation and energy performance improvement, Energy and Buildings. 42
1378 (2010), pp. 290–297. https://doi.org/10.1016/j.enbuild.2009.09.006
1379 [127] Y. Fan, X. Xia, Building retrofit optimization models using notch test data considering
1380 energy performance certificate compliance, Applied Energy. 228 (2018), pp. 2140–
1381 2152. https://doi.org/10.1016/j.apenergy.2018.07.043
1382 [128] N. Elshaboury, M. Marzouk, Optimizing construction and demolition waste
1383 transportation for sustainable construction projects, Engineering, Construction and
1384 Architectural Management. 28 (2020), pp. 2411-2425. https://doi.org/10.1108/ECAM-
1385 08-2020-0636

61
1386 [129] S. Naji, L. Aye, M. Noguchi, Multi-objective optimisations of envelope components for
1387 a prefabricated house in six climate zones, Applied Energy. 282 (2021), pp. 116012.
1388 https://doi.org/10.1016/j.apenergy.2020.116012
1389 [130] R. Chasta, R. Singh, A. Gehlot, R.G. Mishra, S. Choudhury, A smart building
1390 automation system, International Journal of Smart Home. 10 (2016), pp. 91–98.
1391 http://dx.doi.org/10.14257/ijsh.2016.10.8.10
1392 [131] J. Zhu, Y. Shen, Z. Song, D. Zhou, Z. Zhang, A. Kusiak, Data-driven building load
1393 profiling and energy management, Sustainable Cities and Society. 49 (2019), pp.
1394 101587. https://doi.org/10.1016/j.scs.2019.101587
1395 [132] W.H. Allen, A. Rubaai, R. Chawla, Fuzzy neural network-based health monitoring for
1396 HVAC system variable-air-volume unit, IEEE Transactions on Industry Applications.
1397 52 (2015), pp. 2513–2524. https://doi.org/10.1109/TIA.2015.2511160
1398 [133] S. Atis, N. Ekren, Development of an outdoor lighting control system using expert
1399 system, Energy and Buildings. 130 (2016), pp. 773–786.
1400 https://doi.org/10.1016/j.enbuild.2016.08.066
1401 [134] K. Yan, W. Shen, T. Mulumba, A. Afshari, ARX model based fault detection and
1402 diagnosis for chillers using support vector machines, Energy and Buildings. 81 (2014),
1403 pp. 287–295. https://doi.org/10.1016/j.enbuild.2014.05.049
1404 [135] H. Park, D.Y. Park, Comparative analysis on predictability of natural ventilation rate
1405 based on machine learning algorithms, Building and Environment. 195 (2021), pp.
1406 107744. https://doi.org/10.1016/j.buildenv.2021.107744
1407 [136] L. Yang, Z. Nagy, P. Goffin, A. Schlueter, Reinforcement learning for optimal control
1408 of low exergy buildings, Applied Energy. 156 (2015), pp. 577–586.
1409 https://doi.org/10.1016/j.apenergy.2015.07.050
1410 [137] D. Gonçalves, Y. Sheikhnejad, M. Oliveira, N. Martins, One step forward toward smart
1411 city Utopia: Smart building energy management based on adaptive surrogate modelling,
1412 Energy and Buildings. 223 (2020), pp. 110146.
1413 https://doi.org/10.1016/j.enbuild.2020.110146
1414 [138] P. Westermann, M. Welzel, R. Evins, Using a deep temporal convolutional network as
1415 a building energy surrogate model that spans multiple climate zones, Applied Energy.
1416 278 (2020), pp. 115563. https://doi.org/10.1016/j.apenergy.2020.115563
1417 [139] Z. Ding, W. Chen, T. Hu, X. Xu, Evolutionary double attention-based long short-term
1418 memory model for building energy prediction: Case study of a green building, Applied
1419 Energy. 288 (2021), pp. 116660. https://doi.org/10.1016/j.apenergy.2021.116660
1420 [140] Q. Li, R. Long, H. Chen, F. Chen, J. Wang, Visualized analysis of global green
1421 buildings: Development, barriers and future directions, Journal of Cleaner Production.
1422 245 (2020), pp. 118775. https://doi.org/10.1016/j.jclepro.2019.118775.
1423 [141] S.J. Raudys, A.K. Jain, Small sample size effects in statistical pattern recognition:
1424 Recommendations for practitioners, IEEE Transactions on Pattern Analysis and
1425 Machine Intelligence. 13 (1991), pp. 252–264.
1426 https://doi.org/10.1109/ICPR.1990.118138
1427 [142] University of Alabama, Sample Size in Machine Learning and Artificial Intelligence –
1428 Perioperative Data Science, (2021). https://sites.uab.edu/periop-
1429 datascience/2021/06/28/sample-size-in-machine-learning-and-artificial-intelligence/
1430 (accessed November 23, 2021).
1431 [143] I. Balki, A. Amirabadi, J. Levman, A.L. Martel, Z. Emersic, B. Meden, A. Garcia-
1432 Pedrero, S.C. Ramirez, D. Kong, A.R. Moody, P.N. Tyrrell, Sample-Size Determination
1433 Methodologies for Machine Learning in Medical Imaging Research: A Systematic
1434 Review, Canadian Association of Radiologists Journal. 70 (2019), pp. 344–353.
1435 https://doi.org/10.1016/j.carj.2019.06.002.

62
1436 [144] C.N. Egwim, H. Alaka, L.O. Toriola-Coker, H. Balogun, F. Sunmola, Applied artificial
1437 intelligence for predicting construction projects delay, Machine Learning with
1438 Applications. 6 (2021), pp. 100166. https://doi.org/10.1016/j.mlwa.2021.100166.
1439 [145] E. Zarghami, H. Azemati, D. Fatourehchi, M. Karamloo, Customizing well-known
1440 sustainability assessment tools for Iranian residential buildings using Fuzzy Analytic
1441 Hierarchy Process, Building and Environment. 128 (2018), pp. 107–128.
1442 https://doi.org/10.1016/j.buildenv.2017.11.032.
1443 [146] M.R. Hallowell, J.A. Gambatese, Qualitative Research: Application of the Delphi
1444 Method to CEM Research, Journal of Construction Engineering and Management. 136
1445 (2010), pp. 99–107. https://doi.org/10.1061/(ASCE)CO.1943-7862.0000137.
1446 [147] T. Krueger, T. Page, K. Hubacek, L. Smith, K. Hiscock, The role of expert opinion in
1447 environmental modelling, Environmental Modelling & Software. 36 (2012), pp. 4–18.
1448 https://doi.org/10.1016/j.envsoft.2012.01.011
1449 [148] S.N. Sivanandam, S. Sumathi, S.N. Deepa, Introduction to fuzzy logic using MATLAB,
1450 Springer Berlin Heidelberg New York, (2007). ISBN-10 3-540-35780-7
1451 [149] S.O. Abioye, L.O. Oyedele, L. Akanbi, A. Ajayi, J.M. Davila Delgado, M. Bilal, O.O.
1452 Akinade, A. Ahmed, Artificial intelligence in the construction industry: A review of
1453 present status, opportunities and future challenges, Journal of Building Engineering. 44
1454 (2021), pp. 103299. https://doi.org/10.1016/j.jobe.2021.103299.
1455 [150] Y. Duan, J.S. Edwards, Y.K. Dwivedi, Artificial intelligence for decision making in the
1456 era of Big Data – evolution, challenges and research agenda, International Journal of
1457 Information Management. 48 (2019), pp. 63–71.
1458 https://doi.org/10.1016/j.ijinfomgt.2019.01.021.
1459 [151] P. McAleenan, Moral responsibility and action in the use of artificial intelligence in
1460 construction, Proceedings of Institution of Civil Engineers - Management, Procurement
1461 and Law. 173 (2020), pp. 166–174. https://doi.org/10.1680/jmapl.19.00056.
1462 [152] N.V. Chawla, K.W. Bowyer, L.O. Hall, W.P. Kegelmeyer, SMOTE: synthetic minority
1463 over-sampling technique, Journal of Artificial Intelligence Research. 16 (2002), 321–pp.
1464 357. https://doi.org/10.1613/jair.953
1465 [153] I. Guyon, A. Elisseeff, An introduction to variable and feature selection, Journal of
1466 Machine Learning Research. 3 (2003), pp. 1157–1182.
1467 https://psycnet.apa.org/doi/10.1162/153244303322753616
1468 [154] S.G.K. Patro, K.K. Sahu, Normalization: A Preprocessing Stage, ArXiv150306462 Cs.
1469 (2015). http://arxiv.org/abs/1503.06462 (accessed November 24, 2021).
1470 [155] Q. Min, Y. Lu, Z. Liu, C. Su, B. Wang, Machine Learning based Digital Twin
1471 Framework for Production Optimization in Petrochemical Industry, International
1472 Journal of Information Management. 49 (2019), pp. 502–519.
1473 https://doi.org/10.1016/j.ijinfomgt.2019.05.020.
1474 [156] S. Underwood, Blockchain beyond bitcoin, Communication of the ACM. 59 (2016), pp.
1475 15–17. https://doi.org/10.1145/2994581.
1476 [157] M. Crosby, P. Pattanayak, S. Verma, V. Kalyanaraman, Blockchain technology: Beyond
1477 bitcoin, Applied Innovation Review. 2 (2016), pp. 71-90.
1478 [158] A. Shojaei, J. Wang, A. Fenner, Exploring the feasibility of blockchain technology as an
1479 infrastructure for improving built asset sustainability, Built Environment Project and
1480 Asset Management. 10 (2019), pp. 184–199. https://doi.org/10.1108/BEPAM-11-2018-
1481 0142.
1482 [159] B. Panda, S.C. Paul, L.J. Hui, Y.W.D. Tay, M.J. Tan, Additive manufacturing of
1483 geopolymer for sustainable built environment, Journal of Cleaner Production. 167
1484 (2017), pp. 281–288. https://doi.org/10.1016/j.jclepro.2017.08.165

63
1485 [160] R. Parveen, Artificial intelligence in construction industry: Legal issues and regulatory
1486 challenges, International Journal of Civil Engineering and Technology. 9 (2018), pp.
1487 957–62. https://doi.org/10.IJCIET_09_13_096
1488 [161] V. Marinakis, H. Doukas, C. Karakosta, J. Psarras, An integrated system for buildings’
1489 energy-efficient automation: Application in the tertiary sector, Applied Energy. 101
1490 (2013), pp. 6–14. https://doi.org/10.1016/j.apenergy.2012.05.032
1491 [162] W. Wu, R.R. Issa, BIM execution planning in green building projects: LEED as a use
1492 case, Journal of Management in Engineering. 31 (2015), pp. A4014007.
1493 https://doi.org/10.1061/(ASCE)ME.1943-5479.0000314
1494

64