Use of optimized Fuzzy C-Means clustering and supervised classifiers for automobile insurance fraud detection

This paper presents a novel hybrid approach for detecting frauds in automobile insurance claims by applying Genetic Algorithm (GA) based Fuzzy C-Means (FCM) clustering and various supervised classifier models. Initially, a test set is extracted from the original insurance dataset. The remaining train set is subjected to the clustering technique for undersampling after generating some meaningful clusters. The test instances are then segregated into genuine, malicious or suspicious classes after subjecting to the clusters. The genuine and fraudulent records are discarded, while the suspicious cases are further analyzed by four classifiers – Decision Tree (DT), Support Vector Machine (SVM), Group Method of Data Handling (GMDH) and Multi-Layer Perceptron (MLP) individually. The 10-fold cross validation method is used throughout the work for training and validation of the models. The efficacy of the proposed system is illustrated by conducting several experiments on a real world automobile insurance dataset.