Machine Learning Algorithms for Bitcoin Price Prediction

Abstract

Bitcoin has indeed been widely considered as an asset class in recent years, after the asset bubbles of cryptocurrency prices. Because of its extreme volatility, it requires accurate forecasts on which to base financial decisions. Although current research has used machine learning to improve Bitcoin price prediction accuracy, few have looked into the viability of using alternative modelling algorithms on samples with varying data formats and dimensional attributes. To use machine learning approaches to predict Bitcoin price at various frequencies, we first divide Bitcoin price categorized as everyday price and high-frequency price. For Bitcoin exchange rate forecasting, a collection of high-dimension features such as property and network, trade and market, attention, and gold spot price are employed, while good trading features obtained from a bitcoin wallet are being used for five-minute periodic price prediction. The goal of this study is to see how well machine learning models can predict the price of Bitcoin in relation to the US dollar. The learning models are linear regression, random forest, support vector machine, ARIMA, LSTM, and RNN are analyzed in this work. The importance of the sample dimension in machine learning methods is shown in our analysis of bitcoin price prediction. Our research results show that the recurrent neural network outperformed all other models with lower MAPE and RMSPE as 0.3174, 0.8853 respectively.