Customer Churn Analysis
An end-to-end data cleaning, feature engineering and exploratory analysis pipeline on customer churn data. This is the base for my first machine learning model.
- Cleaned and standardized 14 fields: removed duplicates, handled missing values, fixed inconsistent categories.
- Extracted year, month and day features from the join date.
- Found that 27.3% of customers churned, and that churn rates are almost identical across contract type, payment method, tenure and join year. No single variable explains churn.
- Next: logistic regression and tree-based models, evaluated with proper classification metrics.