Anticipating Cryptocurrency Prices Using Machine Learning

Figure 3

Schematic description of Method 1. The training set is composed of features and target (T) pairs, where features are various characteristics of a currency , computed across the days preceding time and the target is the price of at . The features-target pairs are computed for all currencies and all values of included between and . The test set includes features-target pairs for all currencies with trading volume larger than USD at , where the target is the price at time and features are computed in the days preceding .