’MHD25’, ’MHD50’, ’HD25’, ’HD50’ datasets (e.g. ’MC1_att’) after using the PCA algorithm. ’MHD’ simulations are marked with crosses, and ’HD’ simulations with circles. The predicted clusters are marked with blue and orange. The higher the percentage for each PC the higher the relative variance in the dataset that is observed in the direction of the corresponding eigenvector (for the absolute values see Table 3). We use n_clust = 2 in the k-means algorithm. The data is clearly separated into two distinct clusters by distance: 25 pc and 50 pc (see Fig. 3), while MHD and HD labels are mixed.