How ibtesaamaslam/customer-churn-prediction-model works
Main Components
a. Data Source
• Customer-Churn.csv (raw tabular data)
b. Data Ingestion & Preprocessing
• Notebook cells that load CSV with Pandas
• Missing‐value handling, categorical encoding, feature scaling
c. Model Definition & Training
• ANN layers built in TensorFlow/Keras
• Hyperparameter tuning routines (GridSearch/RandomSearch from scikit-learn wrapper)
d. Evaluation & Reporting
• Metrics computation (accuracy, precision, recall, F1‐score, confusion matrix)
• Visualizations with Matplotlib/Seaborn
e. Execution Environment
• Jupyter Notebook (Churn Prediction Model.ipynb)
• Python runtime with dependencies
Relationships & Data Flow
1. The CSV file feeds into the Preprocessing component.
2. The preprocessed feature matrix flows into the Model Definition & Training component.
3. Trained model outputs predictions which feed into the Evaluation & Reporting component.
4. Optional loop: Hyperparameter Tuning sub‐component uses evaluation metrics to adjust model parameters and loops back into training.
Architectural Patterns & Design Principles
• Linear data‐flow pipeline (ETL → Train → Evaluate).
• Modular separation: clearly distinct stages (preprocessing, modeling, evaluation).
• Reproducibility: encapsulated in a single Jupyter Notebook.
Relevant Technologies & Libraries
• Pandas / NumPy (data ingestion & manipulation)
• Scikit‐learn (preprocessing helpers, train/test split, hyperparameter search)
• TensorFlow / Keras (ANN definition & training)
• Matplotlib / Seaborn (visualization)
• Jupyter Notebook (interactive orchestration)