Cryptocurrencies are a hot topic in recent years. This study aims to predict the future closing price of Bitcoin and Ethereum using different combinations of Long Short-Term Memory (LSTM), bidirectional-LSTM (Bi-LSTM), and Gated Recurrent Unit (GRU) with attention mechanisms like Bahdanau and Luong. To achieve this, data from different time scales are taken. The tuning of model’s hyperparameters is done to improve the performance, and it is evaluated using the Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) metrics. The best results were observed for Ethereum in the very short term when using GRU with Bahdanau's attention. Similarly, the best results for Bitcoin were found in the very short term, when using Bi-LSTM with Bahdanau's attention. Overall results of the experiments reveal that the tuning of hyperparameters improves the performance of the model, and the use of attention mechanism on Bi-LSTM and GRU gives a better prediction.
CITATION STYLE
Yazhini, V., Nimal Madhu, M., Premjith, B., & Gopalakrishnan, E. A. (2023). Deep Learning with Attention Mechanism for Cryptocurrency Price Forecasting. In Lecture Notes in Networks and Systems (Vol. 757 LNNS, pp. 471–484). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-99-5166-6_32
Mendeley helps you to discover research relevant for your work.