ENGLISH] Predicting failures and planning ATM maintenance by using data mining technique

Abstract ATM terminals as the most used e-banking channels could be a threat for banks. Some years ago due to a few numbers of ATM, they were taken into account as a competitive advantage; despite most of the time they were out of order. But nowadays their breakdown is a potential threat for banks to lose their customers. On the other hand the amount of data generated by ATM’s maintenance system increases at an unprecedented rate. In this way data mining is introduced as a method for discovering knowledge from these amounts of data in order to predict their failure and disruption. This paper introduces architecture for collecting data generated from ATMs performance and classifying them by predicting their breakdowns. So a sample of 1039 ATM was selected and the needed data were collected. Data were processed by C&R tree and Chaid models in Clementine software. Due to the better result of C&R tree algorithm, it was selected for modeling. By analyzing the results, ATMs were classified into 3 groups and some suggestions were given to maintenance team for planning better