Hybrid predictive modelling for motor insurance claim

Abstract

The objective of this study is to propose a new hybrid model to predict the Malaysia motor insurance claim by estimating the two important components; claim frequency and claim severity. The proposed model are integrating between grey relational analysis and back propagation neural network. We proposed the hybrid model to handle the issue of the insurance data and the complexity of classical statistical technique. Moreover, the classic statistical techniques are incapable of handling huge information in the insurance data. Thus, hybrid model is proposed because it has a high learning ability and capability to learn. Finally, a comparative analysis is carried out to evaluate the predictive model performance between GRABPNN and BPNN. The results produce by MAPE show a small percentage and thus, show that GRABPNN model provides more accurate predictive results compared to BPNN model.