Cryptocurrency Price Prediction Model Using GRU, LSTM and Bi-LSTM Machine Learning Algorithms

Abstract

The rapid rise of cryptocurrencies has indeed created both investment opportunities and forecasting challenges. Accurate predictions of cryptocurrency prices are crucial for traders and financial planners to make informed decisions. Recent studies have employed deep learning techniques, specifically Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), and Bidirectional LSTMs (Bi-LSTM), to analyze the price movements of popular cryptocurrencies. In a comparative analysis, the GRU model consistently outperformed both LSTM and Bi-LSTM models across various cryptocurrencies, demonstrating its effectiveness in modeling financial data. The performance of these models was evaluated using the Mean Absolute Percentage Error (MAPE) after preprocessing, normalizing, training, and evaluating the data. The results indicated that GRU provided the most accurate predictions, with lower MAPE values compared to its counterparts. For instance, a study focusing on Bitcoin, Ethereum, and Litecoin found that the GRU model achieved superior performance metrics, confirming its suitability for cryptocurrency price forecasting. This suggests that traders and investors can rely on GRU-based models for more accurate predictions, ultimately aiding in better investment strategies in the volatile cryptocurrency market.