A Novel Method Based on Support Vector Machines to Classify Bank Transactions

Abstract

Improvements in information technology have contributed to the development of the e-banking industry. Specifically, despite the reduction of bank charges, e-banking is one of the payment methods that, by employing it based on valid theory, can be successful in satisfying customers due to the easiness of access to financial transactions at any time and place with minimum required tools. A mobile device imposes an increasing amount of time, energy and expense in comparison with face-to-face visits. In spite of many benefits this channel has for customers, there are security concerns for service providers and users in the banking sector. Consequently, in this inquiry, the focus is on the role of the support vector machine neural network in the classification of Mellat mobile transactions. To implement the intended procedure, after compiling the information in the preprocessing stage and purification and normalization of data, feature selection is done with the main component analysis algorithm. Then, in post-processing stage, the Neural Network supports the Mobile Banking classification as a safe but fake system. In order to compare the suggested method, we use Bayon floors and multilayer perceptron. The outcomes demonstrate that the support vector machine neural network can fulfill the classification of user's mobile banking transaction with a mean square error of 0.216 and a precision of 94.6% of all data.