Cryptocurrency Price Analysis and Prediction Using LSTM

Abstract

It is difficult to predict cryptocurrency values because of the market's extreme volatility. We suggest a prediction system that makes use of LSTM networks, a potent deep learning technique for identifying temporal patterns in time-series data, in order to solve this. In order to help investors make wise choices, this method is made to forecast the values of eight significant cryptocurrencies, such as BTC, ETH, and BNB. Performance indicators including MSE, MAPE, R2 Score, and RMSE are used to assess the model's efficacy. A comparison with alternative models, such as Gated Recurrent Unit (GRU), linear regression, and conventional time series analysis, is also carried out. thorough testing using a dataset with 8,000 timesteps that were captured at 4-hour intervals shows the LSTM model's better performance, attaining a MAPE of 80% between March 2021 and March 2023. This method offers insightful information that facilitates improved risk management and decision-making in the extremely unpredictable bitcoin market.