How riyadhami13/customer-churn-prediction-pipeline works
This repository is a banking-customer churn analysis, documented through its README and represented by a notebook, a CSV, and a SQLite database. The documented workflow extracts a high-balance inactive-account cohort, evaluates logistic regression and random-forest models, compares score thresholds, and summarizes customers into risk tiers to inform retention decisions. The README documents these stages and results, but no notebook source was available, so internal calls and the roles of the two data files beyond the documented SQL extraction are not asserted.
Main entry points: Churn.ipynb Random forest · churn_banking.db Banking database
Components 12
Customer data
- Banking database
churn_banking.db - Customer CSV
Churn_Modelling.csv - Cohort extraction
Churn.ipynb
Churn modeling
- Churn notebook
Churn.ipynb - Train-test split
Churn.ipynb - Logistic baseline
Churn.ipynb - Random forest
Churn.ipynb - Model evaluation
Churn.ipynb
Retention decisions
- Threshold tradeoff
Churn.ipynb - Risk tiers
Churn.ipynb - Retention recommendation
Churn.ipynb
Other components
- Analyst
Connections 11
- Banking database → Cohort extraction: SQL extraction
README.md - Churn notebook → Cohort extraction: documents workflow
README.md - Cohort extraction → Train-test split: 80/20 split
README.md - Train-test split → Logistic baseline: fits baseline
README.md - Train-test split → Random forest: fits model
README.md - Logistic baseline → Model evaluation: reports metrics
README.md - Random forest → Model evaluation: reports metrics
README.md - Random forest → Threshold tradeoff: scores thresholds
README.md - Random forest → Risk tiers: scores customers
README.md - Threshold tradeoff → Retention recommendation: informs cutoff
README.md - Risk tiers → Retention recommendation: informs targeting
README.md