چكيده به لاتين
Today, in the banking industry, the existence of automated teller machines(ATMs), despite the expansion of e-banking and the nationwide distribution of store terminals, is of great importance. Many customers come to these devices every day to receive banknotes, but the costs of supplying, installing, operating, and supporting ATMs are significant. Considering the banking industry must be profitable, and on the other hand, the supply and maintenance of banknotes and cash is costly for banks, so the cash management of ATMs is of great significance. If the money invested in the device is not commensurate with the money supply period, it can create challenges for banks. For example, if the number of paper money is low, it will reduce customer dissatisfaction, and this will cause the bank to lose the transaction payment, and if the number of paper money in the ATMs is high, then it will waste bank resources and pay cash dividends. In this study, a new method based on clustering and classification of ATMs is presented to optimize their monetization. In the first step, the proposed method calculates the number of suitable clusters for clustering using the voting mechanism. In the second step, with the Kmeans clustering, each sample is placed in the related clusters and its class number is specified. In the third step, the improved artificial neural network is used using a butterfly optimization algorithm(BOA) to classify and identify the type of ATMs. Implementations in MATLAB software show that using 4 clusters has the best results in terms of error index. Experiments show that an increase in the initial population in the butterfly optimization algorithm makes the average error to detect the ATMs type from 0.812 to 0.221, which is about 3.67 times the error reduction. Experiments show that the proposed method has less error than the artificial neural network(ANN), support vector machine(SVM), and decision tree(DT) for the classification of the ATMs.
