Presenting a fraud detection model in online banking systems based on credit card transactions using multiple weighted random forest and quadratic model
Subject Areas :Farzaneh Rahmani 1 , Changiz Valmohammadi 2 , Kiomars Fathi Hafshjani 3
1 - گروه مدیریت فناوری اطلاعات، دانشگاه آزاد اسلامی، واحد تهران جنوب، تهران، ایران
2 -
3 -
Keywords: Fraud detection, online banking, credit card transactions, multiple weighted random forest, quadratic model,
Abstract :
With the increasing growth of online banking, banks and financial institutions are more and more inclined to use this technology and its services. Due to the high volume of transactions, it is practically impossible to manage them by human resources. For this purpose, today, approaches based on data mining have come online with the help of banking. In this article, an efficient model for identifying fraudsters in bank card transactions is presented. The proposed method uses the adjacency matrix, placement of non-valued features using weighting, and random forest aggregation algorithm, in each branch of which, by calculating the weight of each branch, the best branch of the decision maker is selected by calculating the cost of the selection model. It can be It also selects the best forest for decision-making using the multiple quadratic model. Thus, we have tested this method on two data sets, the first one had 14 features and the second one had 20 features, and it has been observed that the model of this research compared to the decision tree, support vector machine, neural network, and normal random forest, which is currently the highest The results have shown improvements over any method. Also, the tests show that none of the mentioned methods were able to predict the OOB error and the normal random forest which is able to predict this error performed much weaker than the proposed model..