Modeling Insurance Fraud Detection Using Ensemble Combining Classification

This paper is a continuation of previous paper where the imbalance dataset problem was solved by applying a proposed novel partitioning-undersampling technique. Then a proposed innovative Insurance Fraud Detection (IFD) models were designed using base-classifiers; Decision Tree, Support Vector Machine and Artificial Neural Network. This paper proposed an innovative insurance fraud detection models by applying ensemble combining classifiers on IFD models designed previously using base-classifiers. Throughout the paper, ten-fold cross validation method of testing is used. Its originality lies in the use of several ensembles combining classifier and comparing between them for choosing the best model. Results from a publicly available automobile insurance fraud detection dataset demonstrate that DTIFD performs slightly better than all proposed models, ensemble combining classifier designed IFD models with high recall but still DTIFD model was the best. The proposed models were applied on another imbalance datasets and compared. Empirical results illustrate that the proposed models gave better results.