Random Rough Subspace based Neural Network Ensemble for Insurance Fraud Detection
In this paper, a random rough subspace based neural network ensemble method is proposed for insurance fraud detection. In this method, rough set reduction is firstly employed to generate a set of reductions which can keep the consistency of data information. Secondly, the reductions are randomly selected to construct a subset of reductions. Thirdly, each of the selected reductions is used to train a neural network classifier based on the insurance data. Finally, the trained neural network classifiers are combined using ensemble strategies. For validation, a real automobile insurance case is used to test the effectiveness and efficiency of our proposed method with two popular evaluation criteria including the percentage correctly classified (PCC) and the receive operating characteristic (ROC) curve. The experimental results show that our proposed model outperforms single classifier and other models used in comparison. The findings of this study reseal that the random rough subspace based neural network ensemble method can provide a faster and more accurate way to find suspicious insurance claims, and it is a promising tool for insurance fraud detection.
insurance fraud detection Rough set Random rough subspace Neural network Ensemble
Wei Xu Shengnan Wang Dailing Zhang Bo Yang
School of Information, Renmin University of China and Key Laboratory of Data Engineering and Knowled School of Information Renmin University of China Beijing, 100872, China
国际会议
昆明、丽江
英文
1276-1280
2011-04-15（万方平台首次上网日期，不代表论文的发表时间）