CHAPTER TWO

LITERATURE REVIEW

2.1 Theoretical Framework

The theoretical framework for this study is based on data mining

methodologies and educational data analytics, which provide a foundation for
extracting meaningful patterns from student data to predict academic performance.
The framework integrates theories from data mining, machine learning, and
educational psychology to establish a comprehensive approach to student
performance evaluation.

2.1.1. Educational Data Mining (EDM) Theory

Educational Data Mining (EDM) is an interdisciplinary field that applies

data mining techniques to educational data to better understand students' learning
behaviors and outcomes. Romero & Ventura define EDM as the process of
transforming raw educational data into useful information that aids decision-
making. This theory emphasizes the use of classification, clustering, and
association rule mining to extract insights about student performance. (Romero &
Ventura 2021)

2.1.2. Constructivist Learning Theory

Piaget's Constructivist Learning Theory suggests that students build

knowledge based on their experiences and interactions. By applying this theory,
data mining models can consider various learning styles and engagement levels to
predict academic success. The integration of EDM with constructivist learning
theory enables a better understanding of how different learning approaches affect
student performance. (Piaget's, 2021)
2.1.3. Machine Learning and Predictive Analytics

Machine learning theories provide the foundation for predictive analytics in

academic performance. Algorithms such as Decision Trees, Naïve Bayes, Random
Forest, and Neural Networks are applied to student data to predict future
performance based on historical data. According to Hastie, Tibshirani, & Friedman
(2020), machine learning techniques help in recognizing patterns that are not
immediately visible through traditional statistical methods.

2.1.4. Tinto’s Student Retention Theory

Tinto’s Student Retention Theory focuses on factors influencing student

retention and dropout rates. This theory is relevant for data mining applications as
it highlights the importance of academic and social integration in student success.
By analyzing attendance records, engagement levels, and grades, data mining
techniques can help institutions identify at-risk students and provide timely
interventions. (Tinto’s, 2022)

2.1.5. Cognitive Load Theory

Sweller’s Cognitive Load Theory explains that excessive cognitive load

negatively affects student performance. Data mining applications can leverage this
theory by identifying students struggling with high cognitive load and
recommending personalized learning strategies to improve performance.
(Sweller’s, 2020)

2.1.6. Data-Driven Decision-Making (DDDM) Theory

DDDM theory emphasizes the importance of data analytics in informed

decision-making. Schools and universities can use data-driven insights from
academic records, demographic data, and behavioral trends to enhance teaching
methodologies and curriculum design.

2.1.7. Conceptual Framework

Based on these theories, the conceptual framework for this study integrates
data mining techniques with student performance metrics to develop an analytical
model that predicts academic outcomes. The framework consists of three key
components:

 Data Collection and Preprocessing: Gathering student-related data (grades,

attendance, participation, socio-economic factors) and cleaning it for
analysis.
 Data Mining and Machine Learning Models: Applying classification
(Decision Trees, Naïve Bayes), regression, and clustering techniques to
identify patterns.
 Predictive Analytics and Decision Support: Generating insights to assist
educators and policymakers in making data-driven interventions for
academic improvement.

2.2 Review of Related Work

The review of related work focuses on prior research and developments in

data mining applications for determining student academic performance. This
section explores various studies that have utilized data mining techniques to
analyze and predict student outcomes.

2.2.1 Application of Data Mining in Education

Several studies have applied data mining techniques in educational settings

to predict student performance. For example, Kotsiantis et al. (2024) developed a
predictive model using classification algorithms such as Decision Trees and Neural
Networks to analyze student performance data. Their study demonstrated the
effectiveness of machine learning techniques in predicting academic outcomes
based on historical data.

2.2.2 Predictive Modeling of Student Performance

Baker and Yacef highlighted the role of predictive modeling in educational

data mining. Their research showed how techniques such as Support Vector
Machines (SVM) and Naïve Bayes classifiers could be used to predict students’
grades and performance trends. They emphasized the importance of feature
selection in improving the accuracy of predictive models. (Baker and Yacef, 2019)

2.2.3 Factors Affecting Academic Performance

Research by Adekoya et al. (2019) investigated factors influencing student

performance, including attendance, participation, and socio-economic background.
Using association rule mining, they identified significant relationships between
student engagement levels and academic success. Their findings underscored the
importance of behavioral and environmental factors in student performance
analysis.

2.2.4 Early Warning Systems for At-Risk Students

Romero et al. (2021) developed an early warning system that used clustering and
classification algorithms to identify students at risk of failing. Their system
allowed educators to intervene early and provide necessary support to struggling
students. This study demonstrated how data mining can enhance student retention
and academic success.

2.2.5 Comparative Analysis of Data Mining Techniques

A recent study by Jayanthi et al. (2021) compared different data mining techniques
such as Random Forest, Decision Trees, and Logistic Regression in predicting
student academic performance. Their results indicated that ensemble methods,
particularly Random Forest, achieved the highest accuracy in academic
predictions.

2.2.6 Implementation of Educational Recommender Systems

Liu et al. (2020) proposed an educational recommender system that utilized

collaborative filtering and clustering techniques to provide personalized learning
recommendations. Their system enhanced student engagement and performance by
tailoring educational resources to individual learning needs.

Data mining is a branch of computer science which deals with the process of
extracting patterns from large data sets by combining methods from statistics and
artificial intelligence with database management. Data mining is seen as an
increasingly important tool by modern business to transform data into business
intelligence giving an informational advantage. It is currently used in a wide range
of profiling practices, such as marketing, surveillance, fraud detection, and
scientific discovery. (Clifton, 2020)

Fuzzy Set Theory is used in applications involving educational assessment

and performance as it is regarded as efficient and effective in uncertain situations
involving performance assessment. It is known that Expert Fuzzy scoring systems
noted (Nolan 2019); help teachers make assessment in less time and with a level of
accuracy that compares favorably to the best teacher examiner. This paper profiled
students from factual and partly behavioral factors. The factual profile content such
as gender, date of birth, race and others like different test results from the college
each semester obtained from the student records. Performance profiling is
dependent upon motivation, attitudes, peer influence, curriculum and by the
continued real-time monitoring of student’s performance using a simple rapid
response system and as (Luan 2022) noted predicts correctly which student may
need some attention or reinforcements in the course of their education. The model
developed helps achieve a measurable student progress monitoring process that
gives results quickly and meets a larger educational goal benefiting stakeholders in
the educational system and the wider community.

Several authors have written on data mining and its application on several
fields of study. Data mining can be applicable in medicine, academics,
engineering, etc. Frawley defined data mining as “The nontrivial extraction of
implicit, previously unknown, and potentially useful information from data”. It
uses machine learning, statistical and visualization techniques to discovery and
present knowledge in a form which is easily comprehensible to humans. Data
Mining evolved from a simple extraction of raw data to an analytical process of
exploring large amount of data in order to cite the common denominators or
patterns. (Frawley, 2020)

Kantardzic further emphasized that manual extraction of patterns from data

has occurred for centuries. Early methods of identifying patterns in data include
Bayes’ theorem and Regression Analysis. The proliferation, ubiquity and
increasing power of computer technology have increased data collection, storage
and manipulations. As data sets have grown in size and complexity, direct hands-
on data analysis has increasingly been augmented with indirect, automatic data
processing. He stressed that data mining has been aided by other discoveries in
computer science, such as neural networks, clustering, genetic algorithms, decision
trees and support vector machines. Data mining is the process of applying these
methods to data with the intention of uncovering hidden patterns. It has been used
for many years by businesses, scientists and governments to sift through volumes
of data such as airline passenger trip records, census data and supermarket scanner
data to produce market research reports. (Kantardzic 2021)

A primary reason for using data mining is to assist in the analysis of

collections of data. Such data are vulnerable to collinearity because of unknown
interrelations. An unavoidable fact of data mining is that the subsets of data being
analyzed may not be representative of the whole domain, and therefore may not
contain examples of certain critical relationships and behaviors that exist across
other parts of the domain. To address this sort of issue, the analysis may be
augmented, using experiment-based and other approaches, such as Choice
Modeling for human-generated data. In these situations, inherent correlations can
be either controlled for, or removed altogether, during the construction of the
experimental design. (Miller, H. and Han, J., 2022)