Cryptocurrency Investments Forecasting Model Using Deep Learning Algorithms

Abstract

The explosive growth of cryptocurrencies has attracted a considerable number of individuals willing to invest, leading to an exponential increase in their market value and trading volume. However, the cryptocurrency market is highly volatile and presents complex datasets where prediction is extremely challenging. Due to this, this article evaluates the implementation of a Prediction System for Cryptocurrency Investments for users using Deep Learning algorithms, aiming to predict the prices of six cryptocurrencies (BTC, ETH, BNB, LTC, XLM, and DOGE). For this, the use of a genetically tuned algorithm with Deep Learning and techniques based on enhanced trees are proposed to compare them. The best result obtained is that the Gated recurrent unit (GRU) model, has an average MAPE of 4%, followed by Convolutional neural networks (CNN) with an average MAPE of 7% and Direct feedforward neural networks (DFNN) of 12%. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.