Micro-insurance Pricing With XGBoost Model Containing Tweedie GLM

Abstract

Microinsurance is a necessary tool that can be widely used to contribute to alleviating poverty. The low income earners’ hard earned wealth or businesses can be protected from the insurable risks. The insurers could consider modelling these products to increase insurance penetration hence increasing the country’s GDP. This paper, I have compared three pricing methods using the same data to draw a conclusion using RMSE as a metric. The results of the project have shown that RMSE of the last model, XGBoost with Tweedie GLM. This approach was taken because of the availability of microinsurance data and the one that is available is heavily zero rated. Tweedie is the most suited generalized linear model since it is a mixture of Poisson- Gamma distribution. The results are so because of benefits such as regularization; parallel processing;handling missing values ;cross validation and effective pruning. All these have been shown in the main project