Edu Performance Predictor
A Streamlit-based web app that predicts student performance using activity data, provides explainability via SHAP, and offers personalized learning recommendations.
Features
- Predicts student final grade using real interaction data from the OULAD dataset
- Built with XGBoost + SHAP for interpretable AI
- Visual SHAP plots to explain individual predictions
- Personalized suggestions to improve performance
- Deployed as an interactive web app with Streamlit
Technologies Used
- Python
- XGBoost
- SHAP
- scikit-learn
- pandas
- Streamlit
- matplotlib
Dataset (OULAD) https://analyse.kmi.open.ac.uk/open-dataset
We use the studentVle.csv from the OULAD dataset.
The CSV is not included in the repo due to GitHub size limits.
To fetch it automatically, run:
bash python download_data.py