Volume 4, Issue 9, September – 2019 International Journal of Innovative Science and Research Technology

ISSN No: -2456-2165 https://doi.org/10.38124/ijisrt/IJISRT19SEP1550

Predicting Consumer Behavior in E-Commerce

Using Recommendation Systems
Nikhil Sanjay Suryawanshi
California, USA.

Abstract:- As e-commerce platforms continue to expand, Keywords:- E-Commerce, Machine Learning, Trust-Based
understanding consumer behavior has become crucial for Filtering, Context-Aware Models, Hybrid Model, Content-
enhancing customer satisfaction and driving business Based Approach, Neural Collaborative Filtering, Sequence-
success. Recommendation systems play a pivotal role in Aware Recommendation Systems, Diversity Optimization.
predicting consumer preferences and delivering
personalized product suggestions. This paper presents an I. INTRODUCTION
extensive literature review on recommendation
techniques, including collaborative filtering, content-based The rapid growth of e-commerce has revolutionized the
approaches, and hybrid models. Notable advancements, retail industry, offering unprecedented opportunities for
such as the use of deep learning, trust-based filtering, and businesses to reach global markets and for consumers to access
context-aware models, are highlighted. Building on these a vast array of products and services [1]. As online transactions
foundations, we propose a novel model that integrates continue to surge, the volume of data generated by consumer
advanced machine learning algorithms with consumer interactions has exploded, presenting both challenges and
behavior analysis to predict preferences more accurately. opportunities for businesses seeking to understand and predict
The expected results suggest that this model will improve consumer behavior [2]. In this context, recommendation
the precision of recommendations, effectively addressing systems have emerged as a critical tool for e-commerce
challenges like data sparsity and evolving user preferences platforms, aiming to enhance user experience, increase sales,
and enhancing overall customer engagement in e- and foster customer loyalty [3].
commerce environments.

Fig 1 Global E-Commerce Statistics

Recommendation systems, also known as recommender The importance of effective recommendation systems in
systems, are intelligent algorithms designed to suggest items or e-commerce cannot be overstated. Studies have shown that
content to users based on their preferences, past behavior, and personalized recommendations can significantly increase
similarities to other users [4]. These systems have become conversion rates, average order value, and customer retention
ubiquitous in e-commerce, playing a pivotal role in [7]. For instance, Amazon.com attributes up to 35% of its
personalized marketing, product discovery, and decision revenue to its recommendation engine [8]. Similarly, Netflix
support [5]. By employing advanced data mining techniques, estimates that its personalized recommendation system saves
machine learning algorithms, and big data analytics, the company $1 billion per year through reduced churn and
recommendation systems can process vast amounts of increased viewer satisfaction [9].
information to generate accurate and timely predictions of
consumer preferences and future behaviors [6].

IJISRT19SEP1550 www.ijisrt.com 806

Volume 4, Issue 9, September – 2019 International Journal of Innovative Science and Research Technology
ISSN No: -2456-2165 https://doi.org/10.38124/ijisrt/IJISRT19SEP1550

Fig 2 An Illustrative Example of Observed Data on a user from a Real e-Commerce Site [14]

However, predicting consumer behavior in the dynamic recommendations and proposed techniques for privacy-
and complex environment of e-commerce presents numerous preserving collaborative filtering.
challenges. These include the cold start problem for new users
or items, the sparsity of user-item interaction data, and the need Salakhutdinov et al. [18] demonstrated the potential of
to balance accuracy with diversity in recommendations [10]. using restricted Boltzmann machines for collaborative filtering,
Moreover, the ever-changing nature of consumer preferences, paving the way for deep learning techniques in
the influence of external factors such as seasonality and trends, recommendation systems. Li et al. [19] explored the challenge
and the potential for recommendation fatigue further of transferring knowledge across different domains to improve
complicate the task of accurate prediction [11]. recommendation accuracy. Tintarev and Masthoff [20]
highlighted the importance of providing explanations for
Recent advancements in artificial intelligence and recommendations to improve user trust and satisfaction.
machine learning have led to the development of more
sophisticated recommendation algorithms. Collaborative Evaluating the effectiveness of recommendation systems
filtering, content-based filtering, and hybrid approaches have in predicting consumer behavior has been a significant area of
been extensively studied and applied in e-commerce contexts research. Herlocker et al. [21] provided a comprehensive
[12]. Deep learning techniques, such as neural collaborative review of evaluation metrics for collaborative filtering systems,
filtering and sequence-aware recommender systems, have discussing their strengths and limitations. McNee et al. [22]
shown promising results in capturing complex user-item highlighted the importance of user-centric evaluation metrics,
interactions and temporal dynamics [13]. arguing that accuracy alone is insufficient for assessing the
quality of recommendations.
Despite these advancements, there remains a significant
gap between academic research and the practical Adomavicius et al. [23] provided a multidimensional
implementation of recommendation systems in e-commerce approach to incorporating contextual information into
[15]. Many businesses struggle to effectively integrate recommendation systems, enhancing their ability to predict
recommendation algorithms into their existing infrastructure, user preferences in different situations. O'Donovan and Smyth
scale them to handle large user bases, and adapt them to [24] introduced the concept of trust in collaborative filtering,
evolving business needs and consumer behaviors [16]. demonstrating how trust metrics can improve recommendation
accuracy. Jamali and Ester [25] proposed a model for
This research paper aims to bridge this gap by exploring incorporating social influence into matrix factorization-based
the current state of recommendation systems in e-commerce, recommendation systems, showing improved performance in
analyzing their effectiveness in predicting consumer behavior, predicting user preferences.
and proposing novel approaches to enhance their performance.
By examining case studies, conducting empirical analyses, and He et al. [26] proposed a neural collaborative filtering
synthesizing findings from various disciplines including framework that employed deep neural networks to model user-
computer science, marketing, and consumer psychology, we item interactions. Their approach demonstrated superior
seek to provide a comprehensive framework for understanding performance compared to traditional matrix factorization
and improving the predictive capabilities of recommendation techniques. Cheng et al. [27] introduced wide & deep learning
systems in e-commerce environments. for recommender systems, combining the memorization of
wide linear models with the generalization of deep neural
II. LITERATURE REVIEW networks. This approach was successfully applied to app
recommendations in Google Play.
As recommendation systems rely heavily on user data,
privacy and security concerns have become increasingly Recognizing the importance of context and sequential
important. Ramakrishnan et al. [17] discussed the risks patterns in user behavior, researchers developed more
associated with inferring sensitive information from sophisticated models. Tang and Wang [28] proposed a self-

IJISRT19SEP1550 www.ijisrt.com 807

Volume 4, Issue 9, September – 2019 International Journal of Innovative Science and Research Technology
ISSN No: -2456-2165 https://doi.org/10.38124/ijisrt/IJISRT19SEP1550

attentive sequential recommendation model that captured both Guo et al. [31] proposed TrustSVD, a trust-based matrix
long-term and short-term user preferences. Liu et al. [29] factorization technique that integrated both explicit and
developed a context-aware sequential recommendation implicit influence of rated items and trusted users. Sedhain et
approach using recurrent neural networks, demonstrating al. [32] proposed a social recommendation method using low-
improved performance in capturing the temporal dynamics of rank matrix factorization to address cold-start scenarios. Zhao
user preferences. Cross-domain recommendation continued to et al. [33] proposed a deep reinforcement learning framework
be an active area of research. Hu et al. [30] proposed a tensor- for page-wise recommendations, optimizing for long-term user
based approach for cross-domain collaborative filtering, engagement.
addressing the data sparsity problem by employing information
from multiple domains.

Table 1 Optimizing for Long-Term user Engagement

Ref. Findings Methods used Dataset Limitations
[34] Utilize hybrid recommender systems Item2Item model for The Entree Chicago The trade-off between
combining collaborative filtering and collaborative filtering Restaurant dataset runtime and accuracy
content-based techniques to analyze and content-based was used for in recommendations.
user interactions and item similarities, filtering. Skip-gram experiments. Contains Limitations of single
enhancing prediction accuracy for algorithm for training restaurant and user recommendation
consumer behavior in E-commerce. item2Vec model. session files. techniques addressed
by hybrid systems.
[35] The COREL framework offers an P(DJDI) to predict Training set: purchase Traditional
effective approach for predicting future customer motivations. data recommendation
purchases. By examining product Hierarchical Bayesian algorithm limitations in
associations and customer preferences discrete choice model Target set: product data an e-commerce
for specific features, this method for preferences. context. Importance of
surpasses conventional customer preferences
Test set: product data.
recommendation algorithms in for product features in
performance. decision-making.
[36] Consumer behavior in e-commerce can Clustering based on user Clickstream data from Insufficient feature
be predicted by analyzing clickstream browsing behavior and a European e- problem affects
data, clustering users based on context clicks. commerce retailer. recommendation
browsing patterns, and employing Collaborative filtering Data was collected for accuracy. New user
collaborative filtering to generate using item category for one month, over 10 and item problems lead
personalized recommendations. neighborhood million users. to poor
formation. recommendations.
[37] Predicting consumer behavior in E- Web Ontology ICS Machine learning -
commerce involves analyzing user data Language (OWL) file. dataset for online retail
through collaborative filtering and Semantic web and used. Over 56 lakh user
clustering algorithms to recommend ontologies data were utilized for
products based on similar user patterns training.
and past behaviors.
[38] Collaborative filtering recommendation Collaborative filtering - Information overload
systems analyze consumer behavior recommendation problem. Need for
characteristics and potential demand to systems. Large data personalized
provide personalized recommendations, analysis and complex recommendations and
effectively addressing information network. services.
overload in e-commerce environments.
[39] Apply the Maximum Coverage Combinatorial solution 30 days of sales data The maximum
algorithm to improve product based on Maximum from a bookstore. 30 Coverage problem is
recommendations. This approach Coverage problem. days of sales data from NP-complete. The
increases the likelihood of customer Greedy algorithm for a hardware store. greedy algorithm may
purchases by emphasizing a varied optimization and not always find optimal
selection of products instead of diversification. solutions.
concentrating solely on popular items.
[40] Recommendation systems can enhance Partial least squares Data was obtained Limited
consumer behavior prediction in e- (PLS) technique for through questionnaires generalizability due to
commerce by analyzing factors like structural equation from 22 respondents. the small interviewee
perceived value, satisfaction, and modeling. An online Respondents were part- number. Common
purchase intent, though they may not survey method for data time students aged 25- method bias in self-
perform predictive tasks satisfactorily. collection. 44 years. reported surveys.

IJISRT19SEP1550 www.ijisrt.com 808

Volume 4, Issue 9, September – 2019 International Journal of Innovative Science and Research Technology
ISSN No: -2456-2165 https://doi.org/10.38124/ijisrt/IJISRT19SEP1550

[41] Implement tailored offers and employ a Lottery mechanism to 120k skin care products Technical
machine learning model to forecast collect willingness-to- from Amazon.com. simplifications may
customers' willingness to pay (WTP) by pay data. Machine Real-time synchronized affect prediction
analyzing their preferences. This learning for predicting product information accuracy. Assumed
approach can boost the efficiency of consumer's willingness with Amazon. constant variance for
recommendation systems in shaping to pay. normalized WTP
buying choices. estimation.
[42] Incorporate elements from the best- Use the most popular Real-world datasets The large number of
selling products and examine how they items for prediction used for experiments. features impacts model
interact to improve the precision of improvement. Analyze efficiency. Lack of
forecasting consumer behavior in e- interactions among detailed information
commerce recommendation systems. popular items for further due to statistical
improvement. features.
[43] Implement recommendation algorithms User interest concept User's personal Overcomes limitations
that consider users' past purchases and tree based on domain information like age, of traditional consumer
preferences, alongside consumer ontology. Multi-agent- occupation, behavior forecasting
interest models. This approach helps based consumer and interests. User methods. Traditional
effectively analyze and anticipate behavior forecasting browsing behavior and methods lack
consumer behavior in e-commerce model. interaction data. collaboration and
settings. adaptability in
predictions.
[44] Analyze user click patterns and apply Prediction method based Data set from Ali Traditional algorithms
statistical probability methods to on probability statistics. Mobile face data sparsity
improve the precision of Utilizes user-clicking Recommendation issues. Rating data is
recommendations. This approach helps behavior data for Competition 2015. often unavailable in e-
overcome the challenges of limited data predictions. Clicking behavior logs commerce.
that conventional suggestion algorithms from 10,000 mobile
encounter in online retail platforms. users.
[45] Utilize Factorization Machines for Factorization Machines Real-world dataset Prediction focuses on
modeling user purchase behavior based model. Feature used for experimental user behavior, not item
on historical data, addressing high- engineering techniques. results. Large-scale recommendations.
dimensional feature engineering to dataset poses modeling Large-scale dataset
enhance prediction accuracy in e- challenges. poses challenges for
commerce recommendation systems. empirical model-
building.
[46] Recommendation systems predict Natural language An initial random No access to a website
consumer behavior by analyzing processing for comment sample of 1,000 for testing
previous buying patterns and user analysis. RV coefficient diverse products. User recommendations.
comments, utilizing techniques like for product similarity comments extracted Scalability issues with
natural language processing and the RV measurement. from Amazon's expanding the system.
coefficient to identify similar products website.
for suggestions.
[47] Implement a semi-supervised approach A semi-supervised A small amount of Limited purchasing
that integrates reliable purchasing method combining purchasing data is used data for small and
information with less precise browsing purchasing and for supervision. Real medium E-commerce
data to improve the accuracy of browsing data. Item- datasets are utilized for sites. Low validity and
predictions in recommendation systems based collaborative extensive experiments. reliability of browsing
for online retail platforms. This method filtering integration for data challenge
aims to enhance the overall predictions. predictions
performance of E-commerce
recommendation engines by combining
different data sources of varying
quality.
[48] Incorporate elements derived from Features based on Real-world datasets -
trending items, combined with temporally popular used for experiments.
conventional statistical measures and items. Comparison with
time-based patterns, to improve the traditional features like The effectiveness of
precision of predicting consumer statistics and the proposed method
behavior in online shopping collaborative filter. was demonstrated on
recommendation systems. the dataset.

IJISRT19SEP1550 www.ijisrt.com 809

Volume 4, Issue 9, September – 2019 International Journal of Innovative Science and Research Technology
ISSN No: -2456-2165 https://doi.org/10.38124/ijisrt/IJISRT19SEP1550

Sarwar et al. [49] implement nearest-neighbor recommender system that relies on consumer product reviews.
collaborative filtering to suggest purchases and predict movie They employ text mining methods to extract valuable insights
ratings. Their work demonstrates that dimensionality reduction from the review comments and then create an ontology to
methods address the challenges of scaling with large datasets convert the extracted information into a format that can be
while maintaining accuracy. Aciar et al. [50] proposed a effectively used by the recommendation system.

Fig 3 Proposed System Architecture

III. PROPOSED SYSTEM which products a user is most likely to prefer. Collaborative
filtering is used to find patterns in the data by identifying users
The architecture represents a system designed to offer with similar behaviors and recommending products based on
personalized product recommendations by processing and those similarities. Alongside this, content-based filtering relies
analyzing user data through multiple layers of modeling and on product attributes and matches items similar to those the
feedback as shown in Figure 3. user has already shown interest in. The Hybrid Model combines
the strengths of both approaches, offering a more
The process begins with the Data Source component, comprehensive and accurate prediction of user preferences by
which gathers all the necessary information required for employing user behavior patterns and product features.
generating recommendations. This data includes user-specific
details, such as browsing history, purchase patterns, and Once the recommendations are generated by the model,
preferences, as well as product-related information like they move to the Ranking Layer Block, where the system
categories, pricing, and ratings. If additional contextual data, organizes and refines the predictions before presenting them to
such as time of purchase or location, is available, it can also be the user. This step involves ranking the recommended items in
incorporated to enhance the recommendations' relevance. This order of relevance to the user. The Personalization component
data, once collected, moves into the Preprocessing stage. of this layer ensures that the results are closely aligned with
individual preferences, while Diversity Optimization works to
In the Preprocessing step, the collected data undergoes avoid overly repetitive suggestions, offering a broader range of
several transformations to prepare it for analysis. This stage is products that might interest the user. This balance of
crucial for cleaning incomplete or erroneous data and personalization and diversity is crucial for maintaining user
performing feature engineering. Feature engineering allows the engagement and introducing them to new products they might
creation of new data features that can improve the accuracy of not have previously considered.
the model, such as deriving engagement scores or calculating
product popularity. During this phase, an interaction matrix, The recommendations are then presented to the user in
which maps user behaviors such as clicks or purchases, is the Output Layer, where the system provides a set of
generated to serve as the foundation for the collaborative Personalized Recommendations. The experience doesn’t end
filtering process. here, however. An essential feature of the system is its
Feedback Loop, which allows for continuous improvement of
The next critical part of the system is the Model Layer, the recommendations. User feedback on the suggestions,
where the actual recommendation models are applied. This whether through explicit actions like ratings or implicit actions
layer employs three primary modeling techniques to predict such as clicks and purchases, is monitored and analyzed. This

IJISRT19SEP1550 www.ijisrt.com 810

Volume 4, Issue 9, September – 2019 International Journal of Innovative Science and Research Technology
ISSN No: -2456-2165 https://doi.org/10.38124/ijisrt/IJISRT19SEP1550

feedback feeds back into the model, which is updated REFERENCES

accordingly. By regularly adjusting the model based on real-
time user feedback, the system becomes more accurate and [1]. Laudon, K. C., & Traver, C. G. (2016). E-commerce:
personalized over time. Business, technology, society (12th ed.). Pearson.
[2]. Chen, J., & Stallaert, J. (2014) “An economic analysis
IV. EXPECTED RESULTS of online advertising using behavioral targeting”, MIS
Quarterly, 38(2), 429-449.
This system architecture is expected to produce highly [3]. Ricci, F., Rokach, L., & Shapira, B. (Eds.). (2015)
accurate predictions of user preferences due to the combined Recommender systems handbook (2nd ed.). Springer,
use of collaborative filtering, content-based filtering, and the 2015.
hybrid model. Each model contributes to better personalization, [4]. Adomavicius, G., & Tuzhilin, A., “Toward the next
with the hybrid model capitalizing on the strengths of both generation of recommender systems: A survey of the
collaborative and content-based methods. Consequently, users state-of-the-art and possible extensions”, IEEE
are likely to receive more relevant recommendations, which Transactions on Knowledge and Data Engineering,
should lead to increased engagement and higher conversion 2005, 17(6), 734-749.
rates. The Ranking Layer Block ensures that these [5]. Schafer, J. B., Konstan, J. A., & Riedl, J., “E-commerce
recommendations are not only personalized but also diverse, recommendation applications”, Data Mining and
preventing recommendation fatigue caused by repetitive Knowledge Discovery, 2001, 5(1-2), 115-153.
suggestions. Diversity in recommendations also increases the [6]. Chen, J., Tao, Y., Wang, H., & Chen, T., “Big data
likelihood that users will discover new items that interest them, based fraud risk management at Alibaba”, Journal of
further improving their experience. Finance and Data Science, 2015, 1(1), 1-10.
[7]. Dias, M. B., Locher, D., Li, M., El-Deredy, W., &
Moreover, the Feedback Loop will contribute to the Lisboa, P. J., “The value of personalised recommender
system’s adaptability and long-term success. As the system systems to e-business: A case study”, In Proceedings of
collects and integrates feedback from users, it will continuously the 2008 ACM Conference on Recommender Systems
evolve to meet their changing preferences. This adaptability (pp. 291-294). ACM.
ensures that the system remains accurate and effective even as [8]. Linden, G., Smith, B., & York, J., “Amazon.com
the product catalog or user base grows, making it scalable and recommendations: Item-to-item collaborative
flexible for different environments. filtering”, IEEE Internet Computing, 2003, 7(1), 76-80.
[9]. Gomez-Uribe, C. A., & Hunt, N., “The Netflix
Another expected result is an increase in overall user recommender system: Algorithms, business value, and
satisfaction and retention. By providing tailored innovation”, ACM Transactions on Management
recommendations and incorporating feedback into model Information Systems, 2015, 6(4), 1-19.
updates, the system can foster a personalized shopping [10]. Rashid, A. M., Albert, I. et al., “Getting to know you:
experience that resonates with individual users. This approach Learning new user preferences in recommender
has the potential to improve sales, as users are more likely to systems”, In Proceedings of the 7th International
act on recommendations that align with their preferences. Conference on Intelligent User Interfaces (pp. 127-
Ultimately, the system will drive greater user engagement, 134). 2002.
satisfaction, and business value. [11]. Yongfeng Zhang, Guokun Lai, Min Zhang, Yi Zhang,
Yiqun Liu, and Shaoping Ma., “Explicit factor models
V. CONCLUSION for explainable recommendation based on phrase-level
sentiment analysis”, In Proceedings of the 37th
This paper has provided a detailed review of international ACM SIGIR conference on Research &
recommendation systems in the context of e-commerce, development in information retrieval (SIGIR '14).
emphasizing their role in predicting consumer behavior. Association for Computing Machinery, New York,
Through an examination of existing approaches, such as NY, USA, 2014, 83–92.
collaborative filtering, deep learning frameworks, and context- https://doi.org/10.1145/2600428.2609579
aware models, we identified key areas where traditional [12]. Burke, Robin. (2002), “ Hybrid Recommender
techniques can be improved. To address these gaps, we Systems: Survey and Experiments”, User Modeling
proposed a machine learning-based recommendation system and User-Adapted Interaction. 12.
that dynamically adapts to user preferences and behaviors. 10.1023/A:1021240730564.
Although the proposed model has yet to be fully implemented, [13]. Zhang, S., Yao, L., & Sun, A., “Deep learning based
we expect it to outperform current systems by providing more recommender system: A survey and new perspectives”,
accurate and personalized recommendations. Future work will ACM Computing Surveys, 2017, 52(1), 1-38.
focus on validating this model with real-world e-commerce [14]. Zhou, Meizi, et al. "Micro behaviors: A new
data and exploring additional enhancements, such as perspective in e-commerce recommender systems."
incorporating cross-domain knowledge and social influence, to Proceedings of the eleventh ACM international
further boost its predictive accuracy. conference on web search and data mining. 2018.
[15]. Jannach, D., Zanker, M., Felfernig, A., & Friedrich, G.
“Recommender systems: An introduction”, Cambridge
University Press, 2010.

IJISRT19SEP1550 www.ijisrt.com 811

Volume 4, Issue 9, September – 2019 International Journal of Innovative Science and Research Technology
ISSN No: -2456-2165 https://doi.org/10.38124/ijisrt/IJISRT19SEP1550

[16]. Xavier Amatriain and Justin Basilico. 2016. “Past, [29]. Liu, Qiang, et al. "Predicting the next location: A
Present, and Future of Recommender Systems: An recurrent model with spatial and temporal contexts."
Industry Perspective”, In Proceedings of the 10th ACM Proceedings of the AAAI conference on artificial
Conference on Recommender Systems (RecSys '16). intelligence. Vol. 30. No. 1. 2016.
Association for Computing Machinery, New York, [30]. Hu, Liang, et al. "Personalized recommendation via
NY, USA, 211–214. cross-domain triadic factorization." Proceedings of the
https://doi.org/10.1145/2959100.2959144 22nd international conference on World Wide Web.
[17]. Ramakrishnan, N., Keller, B. J., Mirza, B. J., Grama, 2013.
A. Y., & Karypis, G., “Privacy risks in recommender [31]. Guo, G., Zhang, J., & Yorke-Smith, N., “TrustSVD:
systems”, IEEE Internet Computing, 2001, 5(6), 54-62. Collaborative filtering with both the explicit and
[18]. Salakhutdinov, Ruslan, Andriy Mnih, and Geoffrey implicit influence of user trust and of item ratings”, In
Hinton. "Restricted Boltzmann machines for Twenty-Ninth AAAI Conference on Artificial
collaborative filtering." Proceedings of the 24th Intelligence, 2015.
international conference on Machine learning. 2007. [32]. Sedhain, Suvash, et al. "Social collaborative filtering
[19]. Li, B., Yang, Q., & Xue, X., “Can movies and books for cold-start recommendations." Proceedings of the
collaborate? cross-domain collaborative filtering for 8th ACM Conference on Recommender systems. 2014.
sparsity reduction”, In Twenty-First International Joint [33]. Zhao, Xiangyu, et al. "Recommendations with negative
Conference on Artificial Intelligence, 2009. feedback via pairwise deep reinforcement learning."
[20]. Tintarev, N., & Masthoff, J., “A survey of explanations Proceedings of the 24th ACM SIGKDD international
in recommender systems”, In 2007 IEEE 23rd conference on knowledge discovery & data mining.
international conference on data engineering workshop 2018.
(pp. 801-810). IEEE. [34]. Nguyen, Sang., “Improving Customer Behaviour
[21]. Herlocker, J. L., Konstan, J. A., Terveen, L. G., & Prediction with the Item2Item model in Recommender
Riedl, J. T., “Evaluating collaborative filtering Systems. Industrial Networks and Intelligent Systems”,
recommender systems”, ACM Transactions on 5. 1-13. 10.4108/eai.19-12-2018.156079.
Information Systems (TOIS), 22(1), 5-53, 2004. [35]. Qiu, Jiangtao, Zhangxi Lin, and Yinghong Li.
[22]. McNee, S. M., Riedl, J., & Konstan, J. A., “ Being "Predicting customer purchase behavior in the e-
accurate is not enough: how accuracy metrics have hurt commerce context." Electronic commerce research 15
recommender systems”, In CHI'06 extended abstracts (2015): 427-452.
on Human factors in computing systems,2006, (pp. [36]. Gupta, Shalini, and Veer Sain Dixit. "A business
1097-1101). intelligent framework to evaluate prediction accuracy
[23]. Adomavicius, G., Mobasher, B., Ricci, F., & Tuzhilin, for E-commerce recommenders." Computational
A., “Context-aware recommender systems”, AI Science and Its Applications–ICCSA 2018: 18th
Magazine, 2011, 32(3), 67-80. International Conference, Melbourne, VIC, Australia,
[24]. O'Donovan, J., & Smyth, B., “Trust in recommender July 2–5, 2018, Proceedings, Part IV 18. Springer
systems”, In Proceedings of the 10th international International Publishing, 2018.
conference on Intelligent user interfaces, 2005, pp. [37]. Acharjee, Sajal, et al. "Decision Support System for
167-174. Online Product Recommendation Service based on
[25]. Jamali, M., & Ester, M., “A matrix factorization Consumer Behavior." International Journal of
technique with trust propagation for recommendation Computer Applications 975 (2017): 8887.
in social networks”, In Proceedings of the fourth ACM [38]. Hongyan, Liu, and Liu Zhenyu. "E-commerce
conference on Recommender systems,2010, pp. 135- consumer behavior information big data mining."
142. International Journal of Database Theory and
[26]. He, X., Liao, L., Zhang, H., Nie, L., Hu, X., & Chua, Application 9.7 (2016): 135-146.
T. S., “Neural collaborative filtering”, In Proceedings [39]. Mikael Hammar, Robin Karlsson, and Bengt J.
of the 26th international conference on world wide Nilsson., “Using maximum coverage to optimize
web, 2017, pp. 173-182. recommendation systems in e-commerce”, In
[27]. Cheng, H. T., Koc, L., Harmsen, J., Shaked, T., Proceedings of the 7th ACM conference on
Chandra, T., Aradhye, H., ... & Shah, H., “Wide & deep Recommender systems (RecSys '13). Association for
learning for recommender systems”, In Proceedings of Computing Machinery, New York, NY, USA, 265–
the 1st workshop on deep learning for recommender 272. https://doi.org/10.1145/2507157.2507169
systems, Association for Computing Machinery, New [40]. Ladislav, Beránek., Václav, Nýdl., Radim, Remeš.
York, NY, USA, 2016, 7–10. "Factors influencing customer repeated purchase
https://doi.org/10.1145/2988450.2988454 behavior in the e-commerce context.", The
[28]. Jiaxi Tang and Ke Wang., “Personalized Top-N International Scientific Conference INPROFORUM
Sequential Recommendation via Convolutional 2015, November 5 - 6, 2015, České Budějovice, 123-
Sequence Embedding”, In Proceedings of the Eleventh 128, ISBN 978-80-7394-536-7.
ACM International Conference on Web Search and [41]. Zhao, Qi, et al. "E-commerce recommendation with
Data Mining (WSDM '18). Association for Computing personalized promotion." Proceedings of the 9th ACM
Machinery, New York, NY, USA, 2018, 565–573. Conference on Recommender Systems. 2015.
https://doi.org/10.1145/3159652.3159656

IJISRT19SEP1550 www.ijisrt.com 812

Volume 4, Issue 9, September – 2019 International Journal of Innovative Science and Research Technology
ISSN No: -2456-2165 https://doi.org/10.38124/ijisrt/IJISRT19SEP1550

[42]. Chen, Chen., Jiakun, Xiao., Chunyan, Hou., Xiaojie,

Yuan. "Improving Purchase Behavior Prediction with
Most Popular Items." IEICE Transactions on
Information and Systems, 100 (2017).:367-370. doi:
10.1587/TRANSINF.2016EDL8169
[43]. Zhang, YanRong, and ZhiJie Zhao. "Study on
Consumer Behavior Predict in E-commerce Based on
Multi-Agent." International Journal of u-and e-Service,
Science and Technology 7.6 (2014): 403-412.
[44]. Jia, Ru, et al. "E-commerce purchase prediction
approach by user behavior data." 2017 international
conference on computer, information and
telecommunication systems (CITS). IEEE, 2017.
[45]. Chen, Chen, et al. "Purchase behavior prediction in e-
commerce with factorization machines." IEICE
TRANSACTIONS on Information and Systems 99.1
(2016): 270-274.
[46]. W. Kamlor and K. Cosh, "Product discovery via
recommendation based on user comments," 2014 6th
International Conference on Knowledge and Smart
Technology (KST), Chonburi, Thailand, 2014, pp. 41-
45, doi: 10.1109/KST.2014.6775391.
[47]. He, Tieke, et al. "Predicting users’ purchasing
behaviors using their browsing history." Databases
Theory and Applications: 26th Australasian Database
Conference, ADC 2015, Melbourne, VIC, Australia,
June 4-7, 2015. Proceedings 26. Springer International
Publishing, 2015.
[48]. Chen, Chen, et al. "Enhancing purchase behavior
prediction with temporally popular items." IEICE
TRANSACTIONS on Information and Systems 100.9
(2017): 2237-2240.
[49]. B. Sarwar, G. Karypis, J. Konstan, and J. Riedl.,
“Analysis of recommendation algorithms for e-
commerce”, In EC’00 Conference Proceedings, 2000
[50]. S. Aciar, D. Zhang, S. Simoff, and J. Debenham.,
“Recommender system based on consumer product
reviews”, In WI’06 Conference Proceedings, pages
719–723, 2006.

IJISRT19SEP1550 www.ijisrt.com 813