Cash Demand Forecasting for Atm Using Neural Networks and Support Vector Regression Algorithms

In this paper two different methods are used to forecast the daily cash demand for automatic teller machines (ATM). The first method is based on flexible artificial neural network (ANN). The generalization properties of this ANN were improved using special adaptive regularization term. The second forecasting method employs the support vector regression (SPR) algorithm. Performed simulation studies and experimental tests showed tolerable forecasting capacities using the both proposed methods. Despite the today's overenthusiastic beliefs about the capabilities of SPR, our investigation showed however that for this application slightly better result can be achieved using forecasting method based on flexible ANN. At this stage the forecasting schema based on flexible ANN is in the implementing phase for intelligent cash management in ATM network.