This project, developed as part of a collaborative group effort, aims to predict the type of life insurance product customers select using demographic, financial, and behavioural data. Key features include age, gender, profession, income level, and transaction history.
An LGBMClassifier model, optimized with custom class weights and stratified cross-validation, handled the dataset’s class imbalance. Predictions were averaged across folds to improve stability.
This approach placed our team 5th in the Anadolu Hayat Emeklilik Datathon 2024, showcasing the effectiveness of boosting algorithms for customer segmentation in life insurance. Further details can be accessed through the project link.
The model effectively identified life insurance products by leveraging diverse customer data. Key drivers included demographic information and transaction behaviour, highlighting their importance in predicting customer preferences.
This project demonstrates the potential of machine learning in enhancing customer targeting strategies in the insurance sector. Future work could explore additional feature engineering and weight optimization to improve predictive accuracy further.