The paper considers modern recurrent neural networks (RNN). Most attention is paid to popular and powerful architectures – long chain of elements of short-term memory (LSTM) and controlled recurrent units (GRU). A software package for forecasting the cost of quotations has been written and a comparison of two methods has been made.

RNN, LSTM, GRU, forecasting the cost of quotes

Mathematical modelling in actual problems of science and technics

Ekhlakov Roman Sergeevich,  ,  orcid.org/0000-0002-7773-3814,  Financial University under the Government of the RF

Sudakov Vladimir Anatolievich,  ,  orcid.org/0000-0002-1658-1941,  KIAM RAS