Predicting Bitcoin Prices by Applying LSTM, ARIMA, SARIMA, GBRP, and GARCH Models

Abstract

Bitcoin is a promising investment asset for the future, offering a viable option for long-term investors. This study seeks to evaluate and compare the effectiveness and performance of several models, including Long Short-Term Memory (LSTM), Autoregressive Integrated Moving Average (ARIMA), Seasonal Autoregressive Integrated Moving Average (SARIMA), Gradient Boosting Regression Process (GBRP), and Generalized Autoregressive Conditional Heteroskedasticity (GARCH), in predicting Bitcoin prices. The dataset spans Bitcoin price data from 2012 to 2024. The findings reveal that each model demonstrates a positive trend in forecasting Bitcoin price movements. Therefore, Bitcoin is a valuable asset for those willing to invest; however, it may not be suitable for novice investors due to its high volatility. This study advances the development of predictive models leveraging machine learning and statistical methodologies, offering critical insights into Bitcoin’s price behavior for informed investment decision-making.