Fig. 2

Machine learning pipeline. A The raw data is preprocessed (preselected and augmented) and then either directly used to train and test the 3 supervised ML algorithms, or compressed via dimensionality reduction methods (PCA or feature extraction). B Illustrations of the 3 algorithms applied to classify the data. Random forests use a majority vote of decision trees, k-nearest neighbors classify the data according to some distance metric, and logistic regression is a statistical method estimating the probability of a data point belonging to a certain class