Forecasting Bitcoin Prices Using LSTM: A Comparative Evaluation with Future Prediction

The decentralized nature and limited supply of Bitcoin have attracted global interest among investors and traders. However, due to its price volatility, accurate predictions of Bitcoin’s future price have become crucial for informed investment decisions. This research aims to investigate the effectiveness of various models in predicting Bitcoin’s future price, with a specific emphasis on Long Short-Term Memory (LSTM). Additionally, Classical Time Series models like Autoregressive Integrated Moving Average (ARIMA) and Seasonal Autoregressive Integrated Moving Average (SARIMA), Random Forest (RF), and XGBoost models are employed for comparative analysis. The LSTM model, a recurrent neural network known for capturing long-term dependencies, is trained on historical Bitcoin price data to evaluate and compare its forecasting capabilities with other established models. Accuracy, robustness, and computational efficiency are assessed using relevant metrics such as mean absolute error (MAE) and Root Mean Squared Error (RMSE). Furthermore, the study extends to forecast Bitcoin’s future price up to 30 days in advance using the LSTM model, providing valuable insights for investors to make informed decisions and optimize their portfolios.