Optimizing risk-prediction for private liability insurance
Problem
Insurance companies use general linear models (GLM) for predicting risk and premiums. These models require a lot of manual work and analysis for every feature.
Solution
To build a machine learning model that can predict the expected number of reported claims and the expected paid loss for calculating the risk for private liability. The implementation of the model will help identify which features have the highest impact and which can be safely excluded to minimize customer burden.
Results
After testing multiple machine learning models, XGBoost has the capability to predict better results from GLM when predicting frequency for private liability.
