ИНТЕЛЛЕКТУАЛЬНЫЕ МЕТОДЫ ОБРАБОТКИ ДАННЫХ ПРИ ПРОГНОЗИРОВАНИИ ОБОРОТА НАЛИЧНЫХ ДЕНЕЖНЫХ СРЕДСТВ В БАНКОМАТАХ КОММЕРЧЕСКИХ БАНКОВ

Научные труды Вольного экономического общества России. 2019. Т. 218. № 4. С. 481–488.

Астраханцева И. А., Кутузова А. С., Astrakhantsev R.

The Article is focused on a problem of forecasting demand for cash money in ATMs of a commercial Bank. The solution of the forecasting problem let us optimize the processes of liquidity management, cash management and ATM service by the collection service. The method of machine learning - neural network-is used to obtain the forecast of cash turnover. The authors made and trained a model of a multilayer perceptron with one hidden layer in the Python 3 programming language using the Keras library, as well as a model of a recurrent neural network. As a result, the forecast of peaks and declines in demand for cash at the Bank's ATMs was obtained. Based on the forecast load management of ATMs ensures minimal maintenance costs and keep money in the ATM. The algorithm can be replicated to the entire ATM network, as well as applied to another commercial banks.

Keywords: банки, ликвидность, машинное обучение, прогнозирование, банкоматы, нейросеть, перцептрон, Рекуррентная нейронная сеть, инкассация, спрос на наличность