Prospects of LSTM Neural Networks Use in Supply Chain Management When Developing a Crypto Currency Rate Forecast

Abstract

Supply chain management is considered as one of the key elements to leveraging a company’s success. Active global crypto currency market growth “overwhelming†national economic systems contributes to the formation of a new type of economic relations which makes supply chain as the crucial factor in development. Despite the fact that at the current time in the world community a single (unified) approach to the legal regulation of the crypto currency market has not yet been developed, crypto currency is considered by many world regulators as a promising tool in the monetary policy of national economies. In this regard, it seems extremely urgent to solve the problem which reveals the features of the studied market development, as well as the possibility of its forecasting for short- and medium-term periods of time. The research subject is the process of economic and mathematical modelling of time series characterizing the bitcoin exchange rate volatility, based on the use of artificial neural networks. The purpose of the work is to search and scientifically substantiate the tools and mechanisms for developing prognostic estimates of the crypto currency market development. The paper considers the task of financial time series trend forecasting using the LSTM neural network for supply chain strategies. The time series composed of the BTC / USD currency pair data is analysed; the period analysed is a day. The authors analysed the neural network architecture, built a neural network model taking into account the heterogeneity and random volatility of the time series, developed and implemented an algorithm for solving the problem in the Python system. For training the neural network, data were used for the period from September 24, 2013, to March 17, 2019 (a total of 2002 data sets). The experiment boils down to that the constructed neural network model is trying to determine the trend of the time series for one next timeframe. The training was conducted with a "teacher". To determine the prediction error, the root-mean-square error (RMSE) was calculated. The results of the study are of practical interest for both government authorities in the field of crypto currency market regulation and for representatives of the business community, who are integrated or planning to integrate into the global crypto currency transaction system.