This project is a Hackathon problem by Analytics Vidhya. The problem statement is to identify the Probability of the existing customer buying a another product. This is a high relevant problem in many industries including the Insurance as it will increase the TOP-LINE of all the companies.
The final solution submitted in the competition is as below.
- Initial Label Encoding of all the Categorical variables
- Feature Engineering of Vintage column to improve the relation between Vintage column and the Target variable
- Implemented CatBoost Classifier
- Implemented LightGBM model using Bayes Optimization
- Comibined both the models using Average of both the predictions.
- The CV score is around 85.8%