An artificial bee colony-based kernel ridge regression for automobile insurance fraud identification

Abstract

Automobile insurance fraud is gradually spreading in its global scope, and mining automobile insurance fraud is of increasing interest to society. There are few studies on the application of Kernel Ridge Regression (KRR) to insurance fraud. KRR has shown its good generalization performance in many fields. The performance of KRR depends on the parameters. KRR requires fewer parameters. In this paper, an Artificial Bee Colony (ABC) algorithm-based Kernel Ridge Regression, called KRR-ABC, is proposed for automobile insurance fraud detection. The global optimization ability of the ABC algorithm is used to optimize the parameter combination of KRR. It can quickly find the global optimal value without traversing all the grid points and improve the generalization ability of the model. The parallel computing is also used to enhance the computing speed of KRR-ABC algorithm. The performance of the KRR-ABC model is evaluated on eight benchmark data sets and compared with other methods. The experiment results show that the KRR-ABC model has faster run time and better generation performance. The KRR-ABC model is applied to detect automobile insurance fraud and the fraud rules are obtained.