عنوان مقاله [English]
Unemployment Insurance (UI) fraud is an important topic in social insurances which is regarded by law as a penal offense. The best way for fraud evaluation is to prevent and control it at early steps, using previous unfolded fraud documents. In this paper, first of all, the fraud control diagram is depicted, then according to the existence of an appropriate database of social security of UI claimants, two modern data mining techniques have been used for data analysis. The Neural Networks and Decision Tree algorithms create models which have been used for evaluating UI benefits fraud. These obtained models will provide an appropriate opportunity for social insurance authorities to have better conceptual understanding and effective combat against all kinds of insurance fraud, clearly just on time. According to several empirical studies, two obtained models have been applied on real data of 15983 records including new and current UI claimants. The accuracy and efficiency of each model have precisely examined. Indicated results show that the Neural Networks model take the first place by 88% of accuracy and the second place is for Decision Tree model by 84% of accuracy. The most important predictors which build each model for fraud pattern recognition, consist of the insured’s previous occupation, paid premium record and age in Neural Networks and branch’s geographic location, gender and the numbers of sponsorships in Decision Tree.