Designing and Explaining a Model for Profitable Automated Teller Machines (ATMs) Predicting and Locating of Tejarat-Bank in Hamadan Based on Security and Theft Prevention Approach using Data Mining Algorithms

Nowadays, use of automated teller machines (ATMs) have revolutionized the whole banking processes. There is no need for anyone to linger in long queues at the bank to receive money due to ATMs existence. ATM Installation at one site isn't risk-free and could be dangerous for banks in terms of security; therefore, bank security can be increased and theft can be prevented by the right and appropriate security locating of the ATM in a city. In this study, we investigate the prediction of the ATMs number required for a bank and their right locating using various data mining methods such as k-nearest neighbor’s algorithm (k-NN), Rough Sets, classification tree algorithm, regression and neural network. The results of this study illustrate that the use of data mining tools can help officials in predicting and locating the required number of ATMs in Hamadan and prevent theft, increase security, and improve police security activities. In addition to considering the profitability of ATMs, the security coefficient of locating each branch is also considered in the data mining models considered in this study. According to the results, it is very important to pay attention to the following indicators in decision making, predicting number and ATMs locating: contract duration, total commission, total transactions and number of ATMs in each district, theft rate in the district, distance to the nearest police station, number of CCTV cameras in the district, population density and number of guards. Also, these indicators will be the basis for predicting the number of profitable ATMs. The support vector machines (SVMs) algorithm with root mean square error (RMSE) of 8% has better estimation capability than other algorithms.