Research On Non-life Insurance Risk Premium Determination Based On Expectile Neural Network

In the field of non-life insurance,risk premium determination is a very important research topic.As a novel risk measure,the risk premium calculated by expectile premium principle can not only better describe the tail characteristics of loss distribution,but also be a consistent risk measure.This thesis discusses the application of expectile in risk premium determination.Expectile neural network model is applied to risk premium calculation for the first time,and expectile regression-neural network nested model is proposed,which can not only describe the nonlinear relationship between variables,but also improve the calculation efficiency and enhance the interpretability of the model.In addition,for the characteristic of unbalanced distribution in non-life insurance data,the probability estimation steps are improved by referring to the Easy Ensemble idea.In view of the characteristic of numerous classification variables,this paper uses the embedding layer structure to deal with them.Finally,the above methods are applied to simulated data and real data.The study shows that the proposed method performs well in risk premium estimation,model interpretability and efficiency.