Bitcoin Price Prediction: A Machine Learning Sample Dimension Approach

The purpose of the paper is to predict Bitcoin prices using various machine learning techniques. Due to its high volatility attribute, accurate price prediction is the need of the hour for sound investment decision-making. At the offset, this study categorizes Bitcoin price by daily and high-frequency price (5-min interval price). For its daily and 5-min interval price prediction, a set of high-dimensional features and fundamental trading features are employed, respectively. Thereafter, we find that statistical methods like Logistic Regression predict daily price with 64.84% accuracy while complex machine learning algorithms like XGBoost predict 5-min interval price with an accuracy level of 59.4%. This work on Bitcoin price prediction recognizes the significance of sample dimensions in machine learning algorithms.