SELECTED WORK / CASE STUDY
Customer Churn Prediction
Machine learning project for identifying telecom customers at risk of churn using enriched customer, service, billing, and zipcode population data.
Project Overview
This project uses the dataset together with a field dictionary and zipcode population enrichment to build a practical churn prediction pipeline.
This project uses the dataset from Kaggle - Telecom Customer Churn.
Python
Random Forest
CatBoost
Pandas
Scikit-learn
Streamlit
Application Demo
Best Model Performance
Accuracy
0.8559
Precision
0.7749
Recall
0.6444
F1-Score
0.7037
ROC-AUC
0.9150
Gini Coef
0.8300