How vnay0704/customer-churn-prediction-system works
This repository is a telecom customer-churn prediction service. The principal workflow accepts JSON through a FastAPI endpoint, prepares and aligns features, scores them with an XGBoost model, and returns a churn probability and thresholded prediction. The notebook is documented as the EDA and training environment. The API source and notebook were not excerpted, so their internal behavior is represented only where the README documents it; model artifact files are not present in the supplied tree.
Main entry points: app.py FastAPI endpoint · predict.py One-hot encoding
Components 7
Prediction API
- FastAPI endpoint
api/app.py
Inference
- One-hot encoding
src/predict.py - Feature alignment
src/predict.py - XGBoost scoring
src/predict.py - Threshold and result
src/predict.py
Model development
- EDA and training
output.ipynb
Other components
- API client
Connections 7
- API client → FastAPI endpoint: submits JSON
readme.md - FastAPI endpoint → One-hot encoding: requests prediction
readme.md - One-hot encoding → Feature alignment: encodes features
src/predict.py - Feature alignment → XGBoost scoring: aligns columns
src/predict.py - XGBoost scoring → Threshold and result: returns probability
src/predict.py - Threshold and result → FastAPI endpoint: returns result
readme.md - FastAPI endpoint → API client: responds with prediction
readme.md