Comparative Analysis of Deep Learning Models for Bitcoin Price Prediction

Abstract

Imagine the hunt to predict Bitcoin&s;s wildly swinging price as a high-stakes competition among four clever computer programs, because investors really need to know where it&s;s headed to make smart plans. Our study pitted these programs—the classic ARIMA, the modern Facebook Prophet, the powerful XGBoost, and the deep-learning LSTM network—against each other to see which could best guess future Bitcoin prices. Using two main report cards, the MAE and RMSE scores, we found that Prophet and ARIMA were neck-and-neck, but the XGBoost model completely missed the mark, proving highly inaccurate with very high error scores. However, the true champion turned out to be the LSTM neural network, which blew the others out of the water by delivering the lowest error scores on both test and training data, essentially making it the most reliable tool for anyone looking to build a winning strategy in the tricky world of crypto trading.