Comparison and Analysis of the Accuracy of Various Machine Learning Algorithms in Bitcoin Price Prediction

Abstract

Based on the dataset of Bitcoin Price dataset, this paper studied Bitcoin price prediction by using support vector machine model, random forest model, neural network model, XGBoost model and LightGBM model. The models were evaluated by MSE, RMSE, MAE, MAPE and R². First, we divided the Bitcoin Price dataset into a training set and a test set according to a ratio of 7:3, with 70 as the training set and 30 as the test set. We take the stock price change (return) as the target variable, the other variables as the input variables, and use the training set to train the model and the test set to test the model. After model comparison, we found that XGBoost's MSE, RMSE, MAE, MAPE and R² are all optimal, and its prediction effect is also the best. The performance of the other four models ranges from good to different, including LightGBM, random Forest, support vector machine and neural network. Among them, the MSE of the neural network is dozens of times that of the other four models, so it performs the worst. The XGBoost model performs well in dealing with high-dimensional sparse data and nonlinear relationships, while LightGBM and random Forest are suitable for dealing with large-scale data. Support vector machines are suitable for dealing with high-dimensional data and nonlinear relationships, while neural networks require more tuning and optimization to take advantage of their advantages. In summary, the research results of this paper can provide value for the prediction of Bitcoin price in the future, and also provide a certain reference for selecting a suitable machine learning model.