The Long Short-Term Memory (LSTM) Model Combines with Technical Analysis to Forecast Cryptocurrency Prices

Authors

  • Dingyu Fu School of Mathematical Sciences, Universiti Sains Malaysia, 11800 USM, Penang, Malaysia.
  • Mohd Tahir Ismail School of Mathematical Sciences, Universiti Sains Malaysia, 11800 USM, Penang, Malaysia.

Abstract

Cryptocurrency has a considerable market value and massive trading volume. Moreover, it is also known for its extreme volatility. Thus, this paper intends to attempt a new approach to forecast cryptocurrency prices by combining the long short-term memory (LSTM) model and technical analysis. The LSTM model has the advantages of a recurrent neural network and solves the gradient disappearance problem that adjusts weights and biases of long- or short-term memory, which is suitable for processing time series problems. Meanwhile, technical analysis is still a critical price trend analytical method. Overall, the results show that the combined methods get a better effect than only using a single price as a feature. Under the same condition, only using price as features for LSTM model accuracy rate is more than 40% for two different error tolerance, but the model accuracy rate will be improved by more than 60% and 90% if traditional technical indicators are combined as features at the best condition. Moreover, the error rate also reduces for the combined approach compared to the single approach.

Downloads

Published

01-08-2023

How to Cite

Fu, D., & Ismail, M. T. (2023). The Long Short-Term Memory (LSTM) Model Combines with Technical Analysis to Forecast Cryptocurrency Prices. MATEMATIKA: Malaysian Journal of Industrial and Applied Mathematics, 149–158. Retrieved from https://matematika.utm.my/index.php/matematika/article/view/1446

Issue

Section

Articles