Neural network technologies for analysis and risk assessment in forecasting the market of industrial financial instruments

Abstract

The issues of risk analysis and assessment in forecasting the dynamics of prices on the stock exchange and the OTC market have always been a difficult task for many researchers and analysts. Successful investment is largely determined by knowledge of the future situation in the financial market. The article deals with the issues of search and development of technical tools for risk analysis and assessment in forecasting the market of derivative financial instruments under conditions of uncertainty. A recurrent network with a long-term short-term memory cell LSTM, which is one of the most powerful prediction models, was chosen as the architecture of neural networks. The development of technical tools for solving this problem was carried out on the basis of the C# programming language. An additional TA-Lib library served as a mathematical base for technical analysis, from which 42 technical analysis indicators were used during the development process, and for working with the LSTM recurrent network, the library Keras.NET. The ability to efficiently link memory and input remote data using these network memory cells provides a dynamic understanding of the data structure over time. The estimates of the accuracy of the forecast for the cost indicators of a derivative financial instrument obtained as a result of computer experiments showed the high efficiency of the constructed neural network model.