Prediction model

Nowadays, the rapid growth of data in organizations has caused managers to look for a way to analyze them. Extracting useful knowledge from aggregation data can lead to appropriate strategic decision-making for the organization. This paper suggests an application of hybrid network based on amount month demand in every ATM device based on transaction mean of 9 months for 1377 devices to obtain customer behavior patterns, to do so, first designed a basic model based on an auto-regressive with exogenous input network (NARX) then, the optimization of the weight and bias of the designed network is made by the genetic algorithm (GA). As a result, finding the weights of the network represents a nonlinear optimization problem that is solved by the genetic algorithm. Paper results show that the NARX-ANFIS Hybrid network using GA for the learning of rules and to optimize the network weights and weights of the network and the fixed threshold can improve the accuracy of the prediction model. Also, classic models are more efficient and increased benefits and lower financing costs and more rational inventory cash control. As well, the designed model can lead to increase benefits and decrease costs in the bank so that, exact forecast and optimal cash upload in ATMs will lead to increase funds on the bank and rise customers and popularity the brand of the bank.