A Comparative Performance Evaluation of Bitcoin Price Prediction Using Machine Learning Techniques

Abstract

The value of bitcoin as a financial asset is rising and due to its extreme volatility, accurate forecasts are necessary to guide investment choices. Adoption, regulatory developments, geopolitical events, and macroeconomic factors all have an impact on its price. Accordingly, a lot of researchers have looked into a variety of factors that influence the value of bitcoin including the trends that underlie its fluctuations, but limited studies have focused on applying various machine learning techniques in this domain. Therefore, the objective of this study is to analyze multiple algorithms used for machine learning regression model in order to identify the system which can estimate bitcoin values most effectively and accurately based on multiple attributes. To forecast the bitcoin price, the dataset has been analyzed and preprocessed meticulously and then multiple machine learning regression models such as XGBoosting, Gradient Boosting Regressor, Hist Gradient Boosting Regressor, Random Forest, Linear Regression, Support Vector Regressor, Neural Network Regressor, Decision Tree, Gaussian Process Regressor, and K-Nearest Neighbors Regressor. The top findings were 99.497 percent (almost 99.5 percent) R-squared (R2), 0.01281 (RMSE), and 0.005755 (MAE) scores employing gradient boosting regressor model.