Cryptocurrency Prediction Using Deep Recurrent Neural Networks

Predicting future Bitcoin prices is crucial for cus-tomers to maximize their profits and minimize their losses. However, this task is challenging because of the complex temporal relationships between Bitcoin-related features. Moreover, exter-nal factors can influence cryptocurrency movement, resulting in unpredictable price fluctuations. To address this problem, deep recurrent neural network (DRNN)-based sequence learner models have been used to learn complex sequential features. In this study, multiple bidirectional versions of LSTM, GRU, and RNN recurrent layers were designed on DRNN models, and their performances were compared for a one-day-ahead Bitcoin price prediction task. The results show that using a convolutional layer with three bidirectional GRU layer-based DRNN model achieves a superior performance, with an average deviation of 3.81% from the actual Bitcoin price which is calculated with Mean Absolute Percentage Error (MAPE) metric.