How ronaldkanyepi/customer-churn-analysis works
Project type and purpose
• This is a data-science/ML pipeline project implemented as a Jupyter Notebook, aimed at analyzing telecom customer data to predict churn. There is no traditional frontend/backend or web service—rather, the "system" is an automated analysis workflow that takes raw data, processes it, trains multiple models, evaluates them, and produces insights (visualizations, feature-importance, selected model).
File structure
• Top-level files:
• Customer Churn Analysis.ipynb: the primary driver of the pipeline (data loading → EDA → modeling → evaluation → interpretation).
• requirements.txt: pins Python dependencies (Pandas, NumPy, scikit-learn, XGBoost, SHAP, LIME, Optuna, TensorFlow/Keras, etc.).
• images/: stores output plots used in the notebook.
• README.md: describes the workflow step by step.
• No submodules or microservices—everything lives in one notebook.
README for additional insights
• The methodology section outlines a clear sequential pipeline with seven stages:
1. Data Loading & Initial Exploration
2. Data Cleaning & Preprocessing
3. Exploratory Data Analysis (EDA)
4. Feature Engineering
5. Model Building (multiple algorithms + hyperparameter tuning)
6. Model Evaluation (AUC-ROC)
7. Feature Importance (SHAP, LIME)
• Data Source: external CSV on S3.
• Outputs: various diagnostic plots and model artifacts.