aboosaleh mohammadsharifi; Kaveh Kahlili-Damghani; farshid abdi; soheila sardar
Abstract
Recently, Bitcoin as the most popular cryptocurrency, has attracted the attention of many investors and economic actors. The cryptocurrency market has experienced a sharp fluctuation, and one of the challenges is to predict future prices. Undoubtedly, creating methods to predict the price of bitcoin ...
Read More
Recently, Bitcoin as the most popular cryptocurrency, has attracted the attention of many investors and economic actors. The cryptocurrency market has experienced a sharp fluctuation, and one of the challenges is to predict future prices. Undoubtedly, creating methods to predict the price of bitcoin is very exciting and has a huge impact on determining the profit and loss from its trading in the future. In this study, in order to predict the price of Bitcoin, a combination of the ARIMA model and three types of deep neural networks including RNN, LSTM, and GRU have been used. The main purpose of this study is to determine the effect of deep learning models on the performance of predicting the future price of Bitcoin. In the proposed model, first, the linear components in the data set are separated using ARIMA and the resulting residues are transferred separately to each of the neural networks. The results show that the ARIMA-GRU model has better results for RMSE and MAPE criteria than other models. Combined models also perform better than the traditional ARIMA model in forecasting.
