Volume 7, Issue 1, January – 2022 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Analysis of Students’Critical Thinking Skills Using

Data Mining Approaches (Survey Based Research)
Dr. M. Pushpalatha (Prof. and HOD) and Raya Bandyopadhyay
Department of Computer Science and Engineering,
SRM Institute of Science and Technology,
Kattankulathur Campus, Chennai – 603203, Tamil Nadu, India.

Abstract:- Our aim in this project, will be to identify Key It is strongly believed and agreed upon in various
Performance Indexes that can define a student's level of instances that data mining techniques can be employed in
comprehension after studying certain classes and how we online based learning systems to read and understand
can apply those Key Performance Indexes in classification academic profiles of students, predict and improve their
or even use them to calculate the success rate of skills, in quality of work, provide training according to academically
universities or even place specific students in areas where demanding needs of the students, and most importantly,
they can succeed. We can also analyse which data mining follow and understand the critical thinking process of the
algorithm gives us the highest accuracy based on our data student. This can be made easier by taking regular students'
and, address some of the open problems we may tests and tracking their attendance patterns, which is why
encounter as we go along, based on existing research relationships can give us insight into what a student's
literature. motivations might be and where a student can be placed to
achieve better school outcomes.
Understanding the learner's in-depth thinking
process after a lesson or series of lessons, will give us more The performance of students in an institution and
information about where the student is lacking or various departments is challenging and depends on a number
whether the skills are lacking, in the event that most of factors, such as a student’s academic performance,
students seem to lack a certain pattern. This shall enable communication skills, capabilities, problem solving and
more fluid methods for students and academics to be disability factors, etc. Work Integrated Learning is a
classified into a system, we can categorize them based on mechanism to enhance performance practice and
class performances or regular assignments, and find a development of a student in their readiness skills, especially
system which shall give us an understanding about the within graduates. The ability to predict student performance
grasp of a particular student in a certain subject and using data mining techniques will assist potential employers
eventually, the group of students performing well in in placing students in an integrated learning career. The
certain subjects can be placed in opportunities which shall research applies data mining techniques to predict if the
enhance their skill sets and help them pick a customized student is critically thinking correctly and if they will pass or
career for them. fail the course. Work Integrated Learning in certain
Universities refers to work-based learning activities at an
The use of multi-phase analysis and cluster analysis approved industry workstation and the experience that
is intended to be based on data on which Key integrate theory and practice. Application of data mining
Performance Indexes will be determined at the end. Based algorithms are helpful to determine information that can be
on these determining Key Performance Indexes, we can used to establish pedagogical basis for taking educational
access important information and, if possible, present it decisions. The concern that whether a student shall flare well
on a working dashboard. or not in the semester involved or shall prove to be below an
average mark, needs to be predicted and factors need to be
I. INTRODUCTION considered accordingly. Predictions are important for
In this moment, there is an overtly increased identifying and assisting in student’s performance for referral.
advancement of an online based education system, where any Classification algorithms using a data mining tool are then
academically involved activity is being followed or tracked applied to the data set for analysis and student performance.
and are being maintained in multiple databases, log files, The patterns to be discovered in the application of
personal profiles, etc. Online based learning is constantly educational data mining techniques will be used to enhance
producing large volumes of data describing the ongoing decision making in student placement in any field, further.
communication between training and teaching, electrical The prediction of student performance in their final
systems and students. In addition, the proliferation of student semesters, mainly will help employability practitioners and
information and behaviours built into the online based student for higher studies for proper progress. The research
learning system is a growing problem as uncontrolled aims to employ data mining algorithms of classification and
information can give erroneous ideas and opinions without clustering techniques to interpret potential and useful insights
providing any clear information. Several studies suggest from the raw data given. The predictions of student
different ways in employing data mining algorithms and performance in a university help in identifying outstanding
techniques in the online based education and learning system. students for industry placement and other higher study
allocations. The marks acquired by the student during the

IJISRT22JAN413 www.ijisrt.com 258

Volume 7, Issue 1, January – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
emester decide their future for allocation in the various email and credit card purchases - have given us new
environments. resources to study our behavior. In particular, smartphones
have emerged as a new tool for collecting data about human
II. RELATED WORK activity, due to its sensory capabilities and ubiquitous
presence. This thesis explores the question of what we can
 Towards Accurate and Fair Prediction of College Success, learn about human behavior in this rich and ubiquitous
Evaluating Different Sources of Student Data: In higher mobile sensor data.
education, predictable analysis can provide information that
can be applied to a variety of stakeholders such as  Data Mining Methods for Detecting the Most Significant
administrators, educators, and students. Different sets of Factors Affecting Students’ Performance: In the last ten
features are often used for different guessing tasks, e.g., years, we have seen a surge in numerous real life, practical
student activity prediction logs and registrar data to predict studies to understand and explore student data
long-term college success. In terms of fairness, the use of methodologically so as to understand and predict their
institutional data consistently undermines the number of performance and fulfillment. Student’s academically flaring
students who were previously disadvantaged than their well involves higher education institutions’ major interests
peers, while LMS data tends to overestimate some of these and enables newer opportunities. Keeping at par with this
groups more often. Combining the two data sources does decision, there is a huge need to oblige the surge and enable
not completely eliminate bias and still leads to high levels methods that study and understand student performance. It
of stigma among disadvantaged groups. This analysis serves contributes to the improvement of the quality of education,
to inform the cost-effective and equitable use of student and, in turn, encourages educational institutions to gain and
data in predictive analytics applications in higher education. gain unparalleled knowledge from the data collected by
their students.
 Predicting Student Performance in Higher Educational
Institutions Using Video Learning Analytics and Data  On the application of data mining algorithms for
Mining Techniques: Technology paired with innovation predicting student performance, a case study: This paper
enable several higher education institutions to employ explores the use of data mining methods, namely
various forms of academic curriculums - video based classification and aggregation, in the data of students
teaching and learning is one such technique. Exploratory graduating with Information Technology degree from a
data analysis of the digital footprints left behind in this mixed university of study. There is a certain part that
online based curriculum enables to completely grasp the employs the decision tree algorithm which is further
efficacy of this type of program. Video-based online associated with prediction of a student’s performance at the
interaction and learning and transformational teaching can final stage of the code that is involved with social variables
enable several improvements in a student’s academic and results obtained from the initial circa (i.e. preparation)
performance. Knowledge gained from the academic of the learning program. The classifier is used as well to
information system, learning management system and search for important insights that shall prove as worthy key
mobile applications were reviewed and analyzed with the performance indicators of the academic fulfillment of a
help of classification algorithms. In addition, data student. The conclusive results only bring forward how
modification and pre-processing techniques were performed highly data mining shall be recommended in involving
to gain more compact features. In addition, genetic research students, academic advisors and university management, to
and partial analysis were performed to further reduce traits. ultimately put these insights into implementation and
henceforth, improve the quality of any educational process
 Developing classifier for the prediction of students’
and eventually each learning experience of each academic.
performance using data mining classification techniques:
Data mining is employed in academically involved  Prediction of Work Integrated Learning Placement Using
institutions to predict a student’s implementation in Data Mining Algorithms: Educational data mining is used
curriculum using classification strategies. These techniques to study data and discover new trends or patterns that can be
are used in student traits to identify logical patterns that can used to gain insights and hence, take certain decisions. The
be used as a basis for speculation. Digital student classification algorithms are employed so as to predict
information being highly accessible and an overtly gain in student’s performance based on academic datasets and their
the coursing power of computer systems make the whole exploratory analysis, and hence, involving an integrated
process a reality. There is a lot of research done in this way system of learning and placements. The decision tree from
to prevent major student failure. Student features collected the forecast shows students who may be ready to be placed
from various sources were previously reviewed, which were in the workplace. The study compares different data mining
later introduced to Put to select features and finally read and techniques to differentiate a student based on the data set
tested. for the semester before the final exam.
 Data Mining and Visualization of Large Human Behavior  Predicting learner’s performance through video
Data Sets: Traditional methods of studying human behavior sequences viewing behavior analysis using educational
such as surveys and hand-to-hand collections are expensive, data-mining: This paper analyzes the methods that students
time-consuming and therefore cannot be easily used on a partake to interact with academically involved instructional
large scale. In recent years a growing number of digital video groups, and their consequences on their academic
social media activities - for example social networking, fulfillments. In this study, auto-emerging video studies were

IJISRT22JAN413 www.ijisrt.com 259

Volume 7, Issue 1, January – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
categorized into multiple pedagogical lists. It focuses on event that most students seem to lack a certain pattern. To
explaining the student's approach to watching the make it easier for students to be organized systematically, we
instructional video, and how they navigate the video can differentiate ourselves based on class performance or
sequence of that video, in order to predict whether the general activities, and find a system that will give us insight
academic can successfully process and fulfill the session into a particular student's performance in a particular subject
conducted through video. Student video clicks are collected and ultimately, a group of students who excel in certain
and sorted. It employs multiple data mining algorithms subjects. The use of multi-phase analysis and cluster analysis
using K-nearest Neighbors and Multilayer Perceptron is intended to be based on data on which KPIs will be
algorithms to predict academic fulfillments of students. This determined at the end. Based on these determining KPIs, we
approach may help teachers to understand how students can access important information and, if possible, present it
view instructional videos. This shall be implemented in any on a working dashboard.
unethical or illegal methides that the student might be
pertaining to while effective education sessions through IV. METHODOLOGIES INVOLVED
videos and also enable instructors to take necessary
preventive steps in order to avoid such situations. The rationalizing in order to conduct an organized
preview based on relational system is to solely find suitable
 Predicting academic success in higher education, procedures for the pre-existing parameters, to bridge any
literature review and best practices: Student’s success shortcomings in the research conducted beforehand and we
stories invite unending opportunities for institutions that find new material to enable the formation of concrete insights
they are involved with and are often considered as the and conclusions. The purpose for an organized preview in the
tipping factor in many a situations for the institution ongoing literature is to solely uphold any dubious statements
themselves. Early detection students who are not doing so that might be put forth during the process. Next, we shall
well or at the edge shall enable to write more favorable define the methodologies and procedures that help drive our
factors to such metrics and might even lead to prevention of research forward and help us in progression towards
such failures. This paper aims to hand hold faculties and conclusions and results. We are hoping this shall also help us
involved personnel to be able to detect such anomalies and define certain boundaries and help us also identify any
predict results of students and hence, be able to favor limitations that might be present within the research.
towards a more successful batch of students if they are
A. Probable factors to be considered:
willing to. As a compliance, the literature has been
revisited, and the forthcoming insights are integrated into a  Relevant questions, descriptive and quantitative both, shall
systematically modelled procedure, in which possible be pertinent to explore the current literature work that helps
decisions and limitations are fully integrated and defined as in understanding and predicting academic performance of
well as controversies. This procedure shall enable educators students.
to fully understand and unleash their potential on a group of  The utmost frequently employed parameters are cumulative
students who can be understood through the various data grade point average and any internal assessments conducted
mining procedures and hence, enable better results. during the progress of semester. The primary rationale as to
why most scholars employ cumulative grade point average
 Adaptive Test Design with a Naive Bayes Framework: is due to its notable contribution to any work or higher
Bayesian image models are frequently employed to derive studies related institution. This parameter shall also be
certain student-based models from specifically related data. qualified as a calculating factor for academic potential of an
Quite a number of predefined models are at our hand institution.
available to model certain Bayesian image models from pre-  Next, the utmost frequently employed parameter is
existing student or academic datasets. These models can population testing and external testing. Academic
even put certain observations and learnings of an individual demographics concentrate on gender, age, family
to test at a critical level. In fact, such data often contain background, and disability. The rationale as to why scholars
deficit values and, in some cases, deficits in excess of the often employ demographics as gender is due to the reason
values shown. It determines the employability of a simple that male and female academic styles of learning and
heuristic choice of test item based on the level of teaching are often extremely of dissimilar methodologies.
uncertainty. The uncertainty of an object is based on its Male academics are known to have extremely efficient and
initial chances of success, too, because of its complexity. optimized methods of grasping the knowledge and having a
The conclusive insights depict that the model’s performance better understanding of some of the topics, on the other
is affected and that heuristic performance definitely leads to hand, females are known well to complete due diligence in
better results. their methodologies and are rarely found to be slagging,
instead they are known to burn the midnight oil quite some
III. PROBLEM DEFINITION nights!
 There have been several approaches that have been made by
It is a challenging task for both the student and their scholars which involve several psychometric parameters of
mentor to understand what field they can excel in, especially an academic personnel. The psychometric parameters that
in the web-based education system. Understanding the are employed are any interest that an academic take in a
learner's in-depth thinking process after a lesson or series of subject, the enthusiasm to grasp knowledge, the amount of
lessons, will give us more information about where the time spent by oneself to study, and family support. These
student is lacking or whether the skills are lacking, in the parameters have been employed to develop a structured

IJISRT22JAN413 www.ijisrt.com 260

Volume 7, Issue 1, January – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
system that is efficient and optimized and simple to time is truly minimal and hence, the time complexity is
implement. While this helps, in assessing an academic’s optimal and even the integration structure is highly
trends of behavior or even their choices and the possible favoured.
reasoning behind them; however, they are not employed as
usual as we would like since they can only give qualitative V. FUTURE SCOPE
insights and in the research world, quantitative insights are
preferred. The inclusion of meta-analysis in forecasting academic
progression of students should rationalize and encourage
B. Algorithms involved: further involvement in the particular field of studies. It shall
 Decision Tree enhance the online based education organization to monitor
Decision Tree is one of the extremely frequent student performance in a structured manner. Although the
methodologies employed to predict academic performance reported findings are based on an existing literature, i.e., data
of students. It is largely employed due to its simpleton ways directly linked to students only, any additional available data
of being able to understand large or smaller datasets and may be included in the analysis, such as educator data or
their structured formation, and eventually predict the parental data that may be relevant and may lead to discovery
parameters as required by the user. Decision tree models are of new features in the study.
effortlessly comprehended and the whole procedure of their
formation can be transformed into conditional, if-then rules. VI. CONCLUSION
The academic performance of the students is mined from Prediction of academic performances are beneficial for
the datasets that are defined in the online based education both learners and faculties involved with the learning in order
system. to make progress in the system and insights that shall help
 Neural Network model the system optimally. This project tries to understand
The Neural Network one of the second most frequently all the methodologies that were involved in the educational
employed methods in mining data from academically system analysis. However, some algorithms improve the
involved backgrounds. One of the advantages of neural performance of categories when used with advanced
network can be listed as, its proactive ability to detect all processing functions such as sorting and re-measuring data,
and any relationships and equations between the predictive but others do not. Little is known about the complete use of
variables and their possibilities. Neural networks are also different data sources for all speculative activities and the
known to have the ability to recognize completely mapped accuracy of your guesses in relation to the diversity of
out relationships between complicated indirect equations different people. Research data makes a limited contribution
between the predictive variables or indices. Henceforth, the to accurate predictions, as, they are often given to study
neural network process is labelled as one of the better participants themselves, and that is why information is often
methods to predict required variable values in a dataset. biased, leading to inaccurate predictions that are clearly
 Naive Bayes inconsistent with real results and therefore, lead to lower
The Naive Bayes algorithm is an eminent option for the accuracy, or false predictions. Maximum scholars used the
scholars to employ in the prediction business. The purpose cumulative grade point average and internal assessments
of the projects employing this methodology is solely to conducted throughout the semesters as base datasets. While in
search for the by far most efficient technique for predicting forecasting techniques, the classification and clustering
academically involved performance of students in a methods are widely used in the area of academically inclined
compare and contrast manner. The literature only depicts data mining. K-Means, while a widely used algorithm proves
that Naïve Bayes involves all and every available variable unfit for our dataset with only 48 percent accuracy and hence,
present in the dataset. Furthermore, each attribute was other algorithms should be explored in order to gain
analyzed exploratively so as to get a clarified understanding maximum accuracy. Among the classification and prediction
of the independence and the quantitative factor of the techniques, Neural Network and Decision Tree are the most
attributes individually. frequently employed methodologies in the mining of
 K-Nearest Neighbor knowledge from academically involved situations.
K-Nearest Neighbor is one of the widespread algorithms Conclusively, among certain red lights there are plenty of
that has been involved widely in a whole lot of projects paths to worth pursuing to reach content results.
with excellent accuracy and good amount of fulfillment.
The K-Nearest Neighbor algorithm’s perspective has REFERENCES
definitely taken minimal amount of time to categorize and
classify academic performances of the students into various [1.] D. Jackson, “The contribution of work-integrated
clusters. It has also shown to provide detailed patterns learning to undergraduate employability skill
among the population progress of academically involved outcomes,” Asia-Pacific J. Coop. Educ., vol. 14, no. 2,
students in the higher standards. pp. 99–115, 2013.
 Support Vector Machine [2.] Zafra and S. Ventura, “Predicting Student Grades in
Support Vector Machine is one of the supervised Learning Management Systems with Multiple Instance
methodologies employed for classification of datasets. A Genetic Programming,” no. Mil, pp. 307–314, 2009.
few projects employ this particular algorithm since its [3.] K. Pal, “Classification Model of Prediction for
proven to be extremely well-suited for smaller scaled Placement of Students,” Int. J. Mod. Educ. Comput.
datasets. Generally, the compilation time and execution Sci., vol. 11, pp. 49–56, 2013.

IJISRT22JAN413 www.ijisrt.com 261

Volume 7, Issue 1, January – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[4.] N. Nghe, P. Janecek, and P. Haddawy, “A comparative
analysis of techniques for predicting academic
performance,” in 37th ASEE/IEEE Frontiers In
Education Conference, 2007, pp. 7–12.
[5.] AZIZ, N. ISMAIL, and F. AHMAD, “MINING
STUDENTS’ACADEMIC PERFORMANCE.,” J.
Theor. Appl. Inf. Technol., vol. 53, no. 3, 2013.
[6.] S. A. Kumar, “EFFICIENCY OF DECISION TREES
IN PREDICTING STUDENT ’ S ACADEMIC
PERFORMANCE,” Comput. Sci. Inf. Technol., vol. 2,
pp. 335–343, 2011.
[7.] S. Milinković and M. Maksimović, “USING
DECISION TREE CLASSIFIER FOR ANALYZING
STUDENTS ’ ACTIVITIES,” JITA, vol. 3, no. 2, pp.
87–95, 2013.
[8.] J. Ghasemian, M. Moallem, and Y. Alipour, “Predicting
students ’ grades using fuzzy non-parametric regression
method and ReliefF- based algorithm,” Adv. Comput.
Sci. an Int. J., vol. 3, no. 2, pp. 43– 51, 2014.
[9.] O. State, “Mining Parent Socio- Economic Factors to
Predict Students’ Academic Performance in Osun State
College of Technology, EsaOke,” Int. J. Eng. Res.
Technol., vol. 2, no. 12, pp. 1677–1683, 2013.
[10.] M’hammedAbdous, W. He, and C.-J. Yen, “Using Data
Mining for Predicting Relationships between Online
Question Theme and Final Grade,” Educ. Technol.
Soc., vol. 15, no. 3, pp. 77–88, 2012.
[11.] Lara, J., Lizcano, D., Martínez, M., Pazos, J., and T.
Riera, A system for knowledge discovery in e-learning
environments within the European Higher Education
Area – Application to student data from Open
University of Madrid, UDIMA. Computers &
Education, 2014. 72(1): p. 23–36.
[12.] Abdous, M., & Yen, C.J. (2010). A predictive study of
learner satisfaction and outcomes in face-toface, satellite
broadcast, and live video-streaming learning
environments. The Internet and Higher Education,
13(4), 248–257.

IJISRT22JAN413 www.ijisrt.com 262