Gated Recurrent Unit Based Deep Learning Model For Bitcoin Price Prediction

Abstract

Over the years, the fluctuations of price of Bitcoin creates many challenges for the data scientists and market investors in predicting the accurate price of bitcoin. In past, many researchers have implemented different models to predict the price of Bitcoin, but still it is required improvement in accurate prediction. In the present study, we proposed the Gated Recurrent Unit (GRU) model to forecast the accurate price of bitcoin. All the models are experimented using the Yahoo Finance dataset, and the performance of the proposed GRU model is explored with six different deep learning models as Additive Prophet, Multiplicative Prophet, Bidirectional Long Short-Term Memory (BiLSTM), CatBooster, Support Vector Regression (SVR), and hybrid GRU 1D-CNN. The model evaluation is performed based on various metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R2 score. From the experimental analysis it is observed that, the performance of the GRU model is outperformed over it's counterparts and produced better result with more prediction reliability.