The Prediction of Short-Term Bitcoin Dollar Rate (BTC/USDT) using Deep and Hybrid Deep Learning Techniques

Abstract

Bitcoin as a digital cryptocurrency interests the scientists substantially in the areas of computer science, cryptography and economics. In this work, we propose to forecast the last price of Bitcoin Dollar rate in short-term or frequent trading transactions known as day-trading. In addition to statistical indicators such as maximum, minimum, and average prices, technical indicators such as Bollinger band (BB), hour-based moving average (MA), Relative Strength Index (RSI) are also evaluated as a feature set. In order to estimate the price of Bitcoin, different deep and hybrid deep learning methodologies are employed, namely convolutional neural networks (CNNs), long short-term memory networks (LSTMs), convolutional long short-term memory networks (ConvLSTM), CNN Long Short-Term Memory Network (CNN-LSTM). Extensive experiment results exhibit that the usage of ConvLSTM hybrid deep learning model is capable to estimate the price of Bitcoin with 2.4076 of MAPE result.