This directory gives the data for the accuracy of SVC models in predicting intuitive descriptors from single ML descriptors.

Data is provided as arrays (for each system, kernel type, cut-off, and model paramater used in the work) giving the balanced accuracy of model predictions using the top 32 ranked ml descriptors - in order. i.e element one of an array gives the balanced accuracy of predictions when learning from the top ranked ml descriptor


_c_meaning in the filename denotes the model paramater c used. p_1 indicates 0.1. For the ROY system, explored data only goes up to c=100
