Student Performance Prediction with Machine Learning

Predicting student academic performance using 7 ML models on 30+ environmental and personal features

---

Overview

A comprehensive machine learning analysis predicting student academic grades based on environmental, demographic, and lifestyle factors. Compares 7 models (from linear regression to deep neural networks) and includes K-Means clustering for outlier detection. Uses the Kaggle Student Performance dataset (395 students, 32 features).

Research Questions

• Can ML predict student performance based on environmental and personal factors?
• Do parental education levels significantly affect student performance?
• How do personal habits and health impact academic grades?

Models Implemented

Traditional ML

Model	Description
Linear Regression	Baseline regression model
Ridge Regression	L2-regularized linear model
Decision Tree	Non-linear tree-based regression
SVR	Support Vector Regression (linear kernel)

Deep Learning

Model	Description
Basic Neural Network	2-layer Dense with ReLU
Improved Neural Network	3-layer with BatchNormalization
Advanced Neural Network	3-layer with BatchNormalization + Dropout

Unsupervised

Model	Description
K-Means Clustering	Outlier detection and student segmentation

Dataset

Source: Kaggle - Student Performance Data Size: 395 students, 32 features

Key Feature Categories:

• Demographics — Age, sex, family size, parental status
• Education — Mother's/Father's education level, school support, study time
• Lifestyle — Free time, going out, alcohol consumption, health status
• Academic — Past failures, absences, travel time

Target: Average grade (mean of G1, G2, G3 grading periods)

Methodology

1. Preprocessing — Categorical to numerical conversion, MinMax normalization
2. EDA — Distribution analysis, correlation heatmap, feature-target scatter plots
3. Outlier Detection — K-Means clustering with distance-based threshold (1.956)
4. Model Training — 80/20 train-test split across all 7 models
5. Evaluation — MAE, MSE, RMSE, R² Score, Explained Variance

Visualizations

• Correlation matrix heatmap
• Target distribution bar chart
• 30-feature scatter plot grid
• K-Means cluster visualization (PCA)
• Per-model prediction plots with training/validation loss curves

Project Structure

StudentPerformanceAI/
├── StudentPerformanceAI.ipynb                    # Main analysis notebook
├── StudentPerformance_Report_SolimanZakaria.pdf  # Full report
├── Model Plots/                                  # Model performance visualizations
├── Plots/                                        # EDA visualizations
└── Student Performance Dataset/
    └── student_data.csv                          # Source dataset

Getting Started

git clone https://github.com/SolyZak/StudentPerformanceAI.git
cd StudentPerformanceAI
pip install pandas numpy scikit-learn tensorflow matplotlib seaborn
jupyter notebook StudentPerformanceAI.ipynb

License

This project is for educational purposes.