amol-gawade01/Dropout-Predictor ? reverse-engineered prompt

Build me a simple student dropout risk prediction app that can take a student’s details and return a risk score, a risk percentage, a risk tier, and the main reasons behind that prediction.

I want the model to use the synthetic student data in this repo, train an XGBoost classifier, and show explainable results with SHAP so it can highlight the biggest risk factors like attendance, academic difficulty, financial stress, or wellbeing support needs. The app should have a clean FastAPI endpoint for sending student info and getting back the prediction in JSON.

Please make it work as a full prototype from data prep to model training to inference, with tests if possible, and keep the code easy to run locally. If you need to check current docs for anything, go ahead and look them up online.