Cryptocurrency Price Prediction Using Parallel Gated Recurrent Units

TL;DR

This work introduces Parallel Gated Recurrent Units (PGRU) for cryptocurrency price prediction, employing two parallel GRU streams—one processing price features and the other processing blockchain-structure features—whose outputs are fused by a feedforward network to predict the next-day price. The approach uses a z-score normalization, a windowed input construction, and compares against an LSTM-based baseline, achieving notable accuracy with a relatively small feature set. Empirical results on Bitcoin data show MAPE values as low as approximately 2.64%–3.24% for suitable window lengths, along with favorable computational cost. The method demonstrates strong potential for accurate, efficient cryptocurrency forecasting and can be extended to other assets or indicators with additional streams.

Abstract

According to the advent of cryptocurrencies and Bitcoin, many investments and businesses are now conducted online through cryptocurrencies. Among them, Bitcoin uses blockchain technology to make transactions secure, transparent, traceable, and immutable. It also exhibits significant price fluctuations and performance, which has attracted substantial attention, especially in financial sectors. Consequently, a wide range of investors and individuals have turned to investing in the cryptocurrency market. One of the most important challenges in economics is price forecasting for future trades. Cryptocurrencies are no exception, and investors are looking for methods to predict prices; various theories and methods have been proposed in this field. This paper presents a new deep model, called \emph{Parallel Gated Recurrent Units} (PGRU), for cryptocurrency price prediction. In this model, recurrent neural networks forecast prices in a parallel and independent way. The parallel networks utilize different inputs, each representing distinct price-related features. Finally, the outputs of the parallel networks are combined by a neural network to forecast the future price of cryptocurrencies. The experimental results indicate that the proposed model achieves mean absolute percentage errors (MAPE) of 3.243% and 2.641% for window lengths 20 and 15, respectively. Our method therefore attains higher accuracy and efficiency with fewer input data and lower computational cost compared to existing methods.

Figures (9)

• Figure 1: Overall steps of the study: data collection, preprocessing, input--output generation, model training, and evaluation.
• Figure 2: Overview of the proposed PGRU architecture. Two parallel GRU-based networks process price features and structural features, respectively. Their outputs are fused by a feedforward network to produce the final price prediction.
• Figure 3: Alternative architecture based on parallel LSTM networks followed by a fusion network. This structure is used as a baseline for comparison with the PGRU model.
• Figure 4: Prediction result and true Bitcoin price with $w=15$ from November 1, 2020 to January 31, 2021 using the PGRU model.
• Figure 5: Absolute differences between prediction and true price with $w=15$ for the same period as Figure \ref{['fig:pgru_w15_pred']}.