How mayanksharma1007/d2c-customer-churn-part-3-churn-prediction-model---model-card works
This repository documents an offline customer-churn modeling workflow, not a serving API: a notebook preprocesses customer data, trains and evaluates candidate models, analyzes thresholds and errors, and saves model and metric artifacts. The README documents the workflow and reports, but no source excerpts or dataset are present, so the graph reflects documented stages and outputs rather than verified implementation calls.
Main entry points: churn_model.ipynb Model Evaluation · error_analysis.md Error Report · model_card.md Model Card · gradient_boosting.pkl Saved Models
Components 14
Modeling Workflow
- Modeling Notebook
churn_model.ipynb - Preprocessing
churn_model.ipynb - Model Training
churn_model.ipynb - Model Evaluation
churn_model.ipynb - Threshold Analysis
churn_model.ipynb - Error Analysis
churn_model.ipynb
Model Artifacts
- Saved Models
Models/gradient_boosting.pkl - Model Comparison
Part_3_outputs/model_comparison.csv - Evaluation Metrics
metrics.json
Analysis Reports
- Error Report
error_analysis.md
Model Documentation
- Model Card
model_card.md
Other components
- Analyst
- Customer Dataset
- Retention Team
Connections 14
- Analyst → Modeling Notebook: runs
- Customer Dataset → Preprocessing: provides data
- Modeling Notebook → Preprocessing: performs
- Preprocessing → Model Training: prepared data
- Model Training → Model Evaluation: trained models
- Model Evaluation → Threshold Analysis: evaluated models
- Threshold Analysis → Error Analysis: selected threshold
- Model Training → Saved Models: saves models
- Model Evaluation → Model Comparison: records comparison
- Model Evaluation → Evaluation Metrics: records metrics
- Error Analysis → Error Report: documents findings
- Analyst → Model Card: documents model
- Model Card → Retention Team: communicates use
- Error Report → Retention Team: communicates risks