Development of Bitcoin Closing Price Prediction Model Using Machine Learning

Bitcoin’s rise as a decentralized cryptocurrency, powered by blockchain, has transformed financial markets by enabling intermediary-free global transactions. Its price volatility, driven by market sentiment, regulatory shifts, and economic trends, challenges traders and investors. This study predicts Bitcoin’s daily closing price trends (up or down) using an ensemble machine learning model combining logistic regression and XGBoost. Using historical price data and technical indicators (SMA_50, SMA_200, RSI) from the model employs hyperparameter tuning, correlation analysis, and RFE for feature selection. Time Series Split cross-validation ensures temporal integrity. The ensemble achieves 90.4% test set accuracy, with precision, recall, and F1-scores above 0.89, outperforming baseline models. Back testing aligns with February 2025 market shifts. This research highlights ensemble learning’s efficacy in volatile cryptocurrency markets, offering traders a robust tool.