Optimization of ATM filling-in with cash

Abstract

This report presents an approach for modeling daily cash demand for all ATMs in the Credit Agricole Bank network in Serbia. The approach is based on time series and regression methods for forecasting an optimal amount of money that should be placed daily in the ATMs in order to meet customers’ demands and mimimize costs of the bank. Three different types of costs were considered: cash freezing costs, transportation costs and insurance costs. The performance of the resulting forecasts were compared with results of the application that bank uses for prediciton of the time and the amount of filling-in for each ATM based on historical data.