Short-Term Prediction of Crypto-Currencies Using Machine Learning

The purpose of this paper is to apply machine learning techniques to predict the movement of cryptocurrencies on an intraday scale and to develop a trading strategy based on the model. A variety of machine learning algorithms like AdaBoost, RandomForest, XGBoost and Neural Networks has been used and their suitability is judged for the task. This work tries to use different labels and unique features including prediction from forecasts of econometric models like GARCH, volume and trade data and their interaction with returns. Finally, a trading strategy based on the model is proposed and back-tested on unseen data