Predictive Models of Development of the Market of Crypto Currency (ARMA, LSTM): Comparative Analysis

Abstract

Today we can see a profound transformation of the traditional world of money and finance. Innovations in the financial sector, new technologies, tools and systems entail serious and profound changes in the usual financial institutions. One of the most important stages of these transformations was the emergence of cryptocurrencies, accompanied by the rapid development of related technologies and the lack of a clear picture of its future landscape.

The lack of a unified approach to the study of the formation of the cryptocurrency rate, the imperfection of the methods for its forecasting impede understanding and objective perception of the prospects for their development.

The relevance of the topic is due to the increasing pace of development of the cryptocurrency market, and the need to create a model capable of predicting the course of major cryptocurrencies.

To solve this problem, a comparative analysis of prognostic models based on the construction of neural networks and the autoregressive-moving average model is proposed. The calculation procedure involved the construction of a recurrent neural network - LSTM, capable of learning long-term dependencies.

As the initial data necessary for the experiment, data on the Bitcoin cryptocurrency rate were used. Econometric processing of the results was based on the integration of descriptive statistics. Based on the results of the study, a predictive model is built to predict the short-term course of bitcoin.

Keywords: cryptocurrency, bitcoin, economic and mathematical methods, forecasting, ARMA modeling.