| # Customer Churn Prediction — Project Documentation |
|
|
| 📄 **Full Document:** [Customer_Churn_Prediction_Model_Document.md](./Customer_Churn_Prediction_Model_Document.md) |
|
|
| ## Overview |
| Comprehensive project document for building a Customer Churn Prediction system using Adaptive Ensemble Machine Learning with Explainable AI. |
|
|
| ## Document Sections |
| 1. **Title** — Project framing and subtitle |
| 2. **Problem Statement** — Business context, technical challenges, and gaps |
| 3. **Idea of Solution** — Stacking ensemble architecture with 5 base models |
| 4. **Objectives** — Primary/secondary goals and success criteria |
| 5. **Literature Review & References** — 21 cited papers spanning 2016–2024 |
| 6. **Dataset Understanding** — Audit of Telco (52 features) and Bank churn datasets |
| 7. **Proposed Methodology** — 7-phase pipeline from preprocessing to CLV scoring |
| 8. **Implementation Strategy** — Tech stack, 4-week timeline, code architecture |
| 9. **Experimental Design** — 5 experiments, 10 metrics, statistical rigor |
| 10. **Result Analysis** — Expected performance, SHAP analysis, business impact |
| 11. **Iterative Improvement** — 6 iterations from feature engineering to production |
|
|
| ## Key Datasets |
| - [Telco Customer Churn](https://hf.co/datasets/aai510-group1/telco-customer-churn) — 7,043 customers, 52 features |
| - [Bank Customer Churn](https://hf.co/datasets/tayaee/bank-customer-churn-prediction) — 12 features |
|
|
| ## Key Papers |
| - Stacking Ensemble (99.28% acc): [arXiv:2408.16284](https://hf.co/papers/2408.16284) |
| - XGBoost Temporal (1st/575 teams): [arXiv:1802.03396](https://hf.co/papers/1802.03396) |
| - Transformer Time-Series (AUC=0.858): [arXiv:2309.14390](https://hf.co/papers/2309.14390) |
|
|
