03 / Explainable AI
Customer Churn Prediction
Combining classification with explainability so a prediction can be understood and acted upon.
Surface which features contributed to an individual model prediction.
Overview
Customer retention teams need more than a score: they need to understand the factors associated with churn. This project explores how to pair predictive classification with model explanations.
How it works
I developed a machine learning classification workflow using customer demographics, service usage and subscription data. A FastAPI backend and Streamlit interface expose predictions and probability scores.
SHAP explanations help identify features that most influence the model output for individual customers, so the result can be interpreted rather than treated as an opaque number.
What mattered
Interpretability supports conversations about a model, but feature contributions are not the same as causal explanations. A useful analytics interface should distinguish predictive association from business causality.
Technologies
Public links will be added when available. The case study above describes the completed work.