Identity Car Insurance Fraud Using Data Mining

This paper aims to compare three models of fraud detection in bodily injury ofcar insurance to select the best method with the least error.Initially, properties were defined as fraud files, and then ۱۲۰ files were selectedfrom the insurance company such that ۲۰ of them were fraud cases. Afterknowing the dataset, fraud files were predicted using the decision tree, supportvector machine (SVM), and the k-nearest neighbors (KNN) algorithm.Comparing the results shows that in terms of total accuracy, KNN provides thebest result with k = ۱۰. Therefore, it can detect fraudulent car insurance fileswith ۱۰۰% accuracy and without error in any class.

Keywords:

Car insurance , Bodily injury , Fraud , Data mining