Predicting Cryptocurrency Price Returns by Using Deep Learning Model of Technical Analysis Indicators

Abstract

Over the last few years, cryptocurrencies have become a new alternative exchange currency for the global economy. Due to the high volatility in the prices of cryptocurrencies, forecasting the price movements is considered a very complicated challenge in the world of finance. Technical analysis indicators are one of the prediction tools which are widely used by analysts. These indicators, which are explored from the historical prices and volumes, might have useful information on price dynamics in the market. Meanwhile, with the new advances in artificial intelligence techniques, like long short-term memory (LSTM), which is able to keep the track of long-term dependencies; there is the extensive application of deep neural networks for predicting nonstationary and nonlinear time series. This study provides a forecasting method for cryptocurrencies by applying an LSTM multi-input neural network to investigate the prediction power of the lags of technical analysis indicators as the inputs to forecast the price returns of the three cryptocurrencies; Bitcoin(BTC), Ethereum (ETH), and Ripple (XRP) that have the highest market capitalization. The results illustrate that the proposed method helps the investors to make more reliable decisions by significantly improving the prediction accuracy against the random walk over the maximum trading time of BTC, ETH, and XRP datasets. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.