Abstarct: The aim of this study is to investigate data mining and the use of its techniques in fraud detection and provide three classification methods. We describe financial classification and the place of data mining in this frxamework. Since the subject of fraud detection is related to classification methods in data mining , we describe three of them that are logistic Regression, Naive Bayes and Bayesian networks and investigate the relation between logistic Regression and Naive Bayes. For evaluating the stated models, some criteria is explained and finally the best model is chosen. Results showed that the logistic regression and Bayesian networks have the same performance and better than the Naive Bayes.