STUDENT PERFORMANCE
ANALYSIS AND PREDICTION
USING DATA MINING TECHNIQUES
Prepared By:
Computer Science & Engineering Department
Course:
Data Warehouse and Data Mining (CSE 06124159)
Academic Year:
2025 - 2026
Date of Submission:
April 16, 2026
Page 1
Abstract
In the modern educational landscape, the ability to predict student performance early in an
academic cycle is crucial for institutional success and student retention. This project explores the
application of data mining techniques—specifically classification and clustering—to analyze a
variety of student attributes including attendance, socioeconomic factors, and previous academic
history. By leveraging algorithms such as Decision Trees and Naive Bayes, we develop a predictive
model that identifies students at risk of underperformance. The results demonstrate that academic
discipline and attendance are the primary predictors of final grades, allowing for timely intervention
by educators.
Table of Contents
1. Introduction 3
2. Literature Review & Data Collection 4
3. Data Preprocessing & ETL 5
4. System Methodology 6
5. Implementation & Results 7
6. Data Visualization 8
7. Conclusion & References 9
Page 2
Chapter 1: Introduction
1.1 Background
Educational Data Mining (EDM) is an emerging discipline focused on developing methods for
exploring unique types of data that come from educational settings. As databases in universities
grow, traditional manual analysis becomes impossible.
1.2 Problem Statement
Currently, academic performance is evaluated only after examinations are completed. This reactive
approach prevents instructors from assisting struggling students during the semester. There is a clear
need for a proactive system that utilizes historical data to forecast future outcomes.
1.3 Objectives
• To design a data warehouse architecture specifically for academic records.
• To implement classification algorithms to categorize student performance into "Excellent",
"Average", and "At-Risk".
• To evaluate the impact of non-academic factors (e.g., internet access, study time) on final
results.
Page 3
Chapter 2: Literature Review & Data Collection
2.1 Literature Review
Previous studies in the field of EDM have shown that student performance is not solely dependent
on intelligence but is a multivariate problem. Researchers like Romero et al. (2013) successfully
used Association Rule Mining to find patterns between student forum participation and final grades.
2.2 Data Source
The dataset used for this project comprises 500 records of undergraduate students. It includes both
numerical and categorical data types to ensure a robust analysis.
2.3 Data Attributes
Attribute Type Description
Attendance_Rate Numeric Percentage of classes attended
Study_Hours Numeric Average weekly self-study time
Previous_GPA Numeric CGPA from the prior semester
Parental_Education Categorical Highest degree of parents
Page 4
Chapter 3: Data Preprocessing & ETL
Before the data can be fed into a mining algorithm, it must undergo a rigorous cleaning and
transformation process known as ETL (Extract, Transform, Load).
3.1 Data Cleaning
Missing values for 'Study_Hours' were replaced using the median value to avoid skewing the
distribution. Outliers in exam scores were investigated and removed if they resulted from entry
errors.
3.2 Data Transformation
Categorical variables such as "Parental_Education" were encoded using One-Hot Encoding.
Continuous variables were normalized to a range of [0, 1] using Min-Max scaling to prevent
attributes with larger ranges from dominating the model.
3.3 Feature Selection
Using a correlation matrix, we identified that 'Attendance' and 'Previous_GPA' had the highest
coefficient relative to the target variable 'Final_Grade'.
Page 5
Chapter 4: System Methodology
The project follows the standard CRISP-DM (Cross-Industry Standard Process for Data Mining)
methodology.
4.1 Workflow Architecture
[Flowchart: Raw Data -> Cleaning -> Feature Selection -> Model Training ->
4.2 Algorithms
Decision Tree (J48/C4.5): We chose this forPrediction]
its interpretability. It generates a tree structure where
each internal node represents a test on an attribute, allowing teachers to see exactly *why* a student
is predicted to fail.
Naive Bayes: Used as a baseline classifier due to its computational efficiency and performance on
small datasets.
Page 6
Chapter 5: Implementation & Results
The implementation was carried out using Python 3.9 and the Scikit-Learn library. The dataset was
split into 80% training data and 20% testing data.
5.1 Model Performance
Algorithm Accuracy (%) Precision Recall
Decision Tree 84.5% 0.82 0.85
Naive Bayes 79.2% 0.76 0.78
5.2 Analysis
The Decision Tree outperformed Naive Bayes, particularly in identifying students in the "At-Risk"
category. The tree showed that students with attendance below 75% were 3 times more likely to fail
regardless of their previous GPA.
Page 7
Chapter 6: Data Visualization
Visual representation of data helps in identifying trends that are not immediately obvious in tabular
format.
6.1 Study Hours vs. Final Grade
[Scatter Plot showing a positive linear correlation between study hours and
final grades]
6.2 Grade Distribution by Attendance
The following visualization depicts how the "Excellent" grade bracket is dominated by students
with over 90% attendance, confirming the model's primary feature importance.
Page 8
Chapter 7: Conclusion & References
7.1 Conclusion
This project successfully demonstrates that student performance can be predicted with significant
accuracy using data mining. The findings highlight that behavioral factors like attendance are just as
critical as academic history. Implementing this model in a real university setting could provide a
vital early-warning system for academic advisors.
7.2 Future Work
Future iterations will include social media sentiment analysis and extracurricular participation
levels to provide a more holistic view of the student's life and its impact on their studies.
References
1. Han, J., Kamber, M., & Pei, J. (2011). Data Mining: Concepts and Techniques. Morgan
Kaufmann.
2. Romero, C., & Ventura, S. (2013). Educational Data Mining: A Survey from 1995 to 2005.
Expert Systems with Applications.
3. Course Material: CSE 06124159 - Data Warehouse and Data Mining.
Page 9