🎓 Student Performance Prediction

This project uses a dataset of student performance to predict overall academic achievement using a linear regression model.

---

📊 Model Evaluation & Insights

🎯 Objective

This project aims to predict students' average academic performance based on demographic and support-related features, such as:

• Gender
• Lunch type (used as a socioeconomic proxy)
• Test preparation course completion

🧪 Model Metrics

Using Linear Regression, the model yielded the following evaluation metrics:

• Mean Absolute Error (MAE): 11.09
• R-squared (R²): 0.074

These results indicate that the model’s predictions deviate from actual average scores by approximately 11 points, and it explains only about 7.4% of the variance in student performance.

📈 Correlation Analysis

A correlation heatmap was used to explore relationships between variables:

• Subject scores (math, reading, writing) showed strong mutual correlations (0.80–0.95), which is expected.
• Test preparation course completion had a weak-to-moderate positive correlation with scores (up to ~0.31).
• Lunch type showed a similar moderate effect (~0.35).
• Gender had a slight negative correlation with reading and writing scores (~ -0.24 to -0.30).

💡 Insights

• 📚 Test preparation seems to positively impact scores, but the effect is limited.
• 🍱 Lunch type (as a socioeconomic indicator) also correlates modestly with performance.
• 🚻 Gender appears to have a smaller effect, especially in reading and writing.
• 📉 Overall, the model’s low R² suggests that background features alone are not strong predictors of student performance.
• For stronger prediction models, including past academic results, attendance records, or engagement metrics could significantly improve performance.

---