Next Day Bitcoin Price Prediction: Performance Comparison of Various Statistical and Machine Learning Algorithms

Abstract

Trading in Digital currency is an opportunity for an alternate investment option and getting popularity day by day. Bitcoin is one of the most popular digital currencies based on technology implementation. Though it operates free from any central control, still many investors trade in bitcoin and also contribute to the economy. The objective of this research paper is to implement and analyse five different statistical and machine learning algorithms in next day bitcoin price prediction. The different algorithms implemented in this work are Random Forest, Support Vector Regression (SVR), Ridge regression, Lasso regression and Long Short Term Memory (LSTM) models. The data used in this work is the daily traded data for the period from November 2021 to February 2023. From the experiments, it is concluded that LSTM model and Lasso regression predict the same value for the next day bitcoin price with an accuracy of 97.88%followed by Random Forest and Support vector model with accuracy of 94.65% and 94.40% respectively. However, the highest accuracy is observed by the Ridge regression which is 98.02%.