Challenges
- Data in batches
- Data change during analysis resulting in frequent rework
- Variable recoding towards the end of modeling work
Solution
- Data processing and analysis carried out in python
- Scorecard with reason code for LR model
- For machine learning method, XGBoost model showed better performance than Random Forest model
- Score for ML model was generated
- Score interpretation using LIME (Local Interpretable Model Agnostic Explanation)
