How chaitanya-g53/customer-churn-prediction-xgboost works
This repository trains a bank-customer churn classifier and serves interactive predictions through a Streamlit dashboard backed by a local Flask endpoint. The principal workflow is customer-profile entry → HTTP prediction request → feature encoding and XGBoost inference → risk assessment and dashboard display. The README documents the notebook’s exploration, preprocessing, training, evaluation, and feature-importance stages; those details are documented rather than verified from notebook source here.
Main entry points: app.py Prediction Endpoint · Customer-Churn-Prediction-XGBOOST.pkl Trained Model · CustomerChrun_XGBoost.ipynb Model Workflow
Components 10
Model Development
- Model Workflow
CustomerChrun_XGBoost.ipynb - Trained Model
Customer-Churn-Prediction-XGBOOST.pkl
Prediction Service
- Prediction Endpoint
app.py - Feature Encoding
app.py - Model Inference
app.py
Customer Dashboard
- Profile Inputs
app.py - Prediction Client
app.py - Risk Assessment
app.py - Dashboard Results
app.py
Other components
- Dashboard User
Connections 13
- Dashboard User → Profile Inputs: enters profile
- Model Workflow → Trained Model: serializes model
- Trained Model → Model Inference: loaded for inference
- Profile Inputs → Prediction Client: builds payload
- Prediction Client → Prediction Endpoint: POSTs prediction
- Prediction Endpoint → Feature Encoding: encodes features
- Prediction Endpoint → Model Inference: calls model
- Model Inference → Prediction Endpoint: returns prediction
- Prediction Endpoint → Prediction Client: returns JSON
- Prediction Client → Risk Assessment: derives risk
- Risk Assessment → Dashboard Results: supplies display values
- Profile Inputs → Dashboard Results: supplies profile
- Dashboard User → Dashboard Results: views results