Student Performance Analysis and Prediction - Data Science Project

🎓 Project Overview

Achieved 100% Accuracy (R² = 1.0000) predicting student CGPA using Linear Regression on 1,193 student records. Model discovered academic progress perfectly determines current performance.

📊 Dataset

• Source: Kaggle
• Size: 1,193 students (954 train + 239 test)
• Features: 24 total → 5 selected
• Target: current_cgpa (0-4.0 scale)

🚀 Key Results

R² Score: 1.0000 (100% variance explained) ✅ RMSE: 0.0000 CGPA points (perfect) ✅ MAE: 0.0000 CGPA points (perfect) ✅ Train=Test: Perfect generalization ✅

🛠️ Techniques Applied

Data Cleaning: Missing value imputation, binary encoding, outlier removal (3×IQR), StandardScaler
Feature Engineering: academic_progress = current_cgpa - prev_sgpa, correlation selection
Visualization: Correlation heatmap, actual vs predicted scatter, residual plot, feature importance

🔬 Key Findings

Rank	Feature	Coefficient	Impact
1	academic_progress	+1.000	PERFECT PREDICTOR
2	prev_sgpa	+0.829	Strong baseline
3-5	Others	~0.000	Negligible

Model Equation: CGPA = 2.697 + 1.000×(academic_progress) + 0.829×(prev_sgpa)

## 🏗️ **Project Structure**
student-performance-prediction/
├── data/Students_Performance_data_set.xlsx
├── notebooks/01_data_cleaning.ipynb
├── notebooks/02_eda.ipynb
├── notebooks/03_model_training.ipynb
├── src/data_preprocessing.py
├── src/model_training.py
├── models/best_model.pkl
├── reports/project_report.md
└── results/linear_regression_results.png

📋 Requirements pandas==2.0.3 numpy==1.24.3 scikit-learn==1.3.0 matplotlib==3.7.2 seaborn==0.12.2 jupyter==1.0.0

💡 Key Insights Academic progress is mathematically deterministic (1:1 relationship)

Previous SGPA sets performance baseline

Attendance/scholarship have zero direct impact

Perfect linear relationship - no randomness

Production-ready model with zero prediction error

📬 Contact Maneesh Kumar - Computer Science Student Shri Mata Vaishno Devi University (SMVDU) Email: mchaudhary2817@gmail.com LinkedIn: linkedin.com/in/yourprofile