Recent Projects
Intra-day Bitcoin Price Prediction with Machine Learning and Ensemble Methods
This project compares the performance among five basic machine learning models to predict the intra-day five-minute average Bitcoin prices from August 5th to August 31st, 2021 on Coinbase and attempts to use ensemble methods (1) stacking, (2) bagging and (3) random forest to improve forecasting performance. The forecast evaluation criteria entail forecast accuracy, directional accuracy andprofitability. The results find that the random walk model outperforms other models and ensemblemethods in the forecast and directional accuracy. However, ensemble methods are helpful to increasethe profitability of basic machine learning models.