Figure 12.

Feature significance of each SVM model using the ML-insights python package. (a) shows the importance of each feature when only volumetric features are used. (b) shows the importance of each feature when only network properties are used. (c) shows the importance of each feature when volumetric and network properties are used. It is interesting to see that in (c), the most important features are a mixture of volumetric features and network properties. For (a) and (c), linear scale was used, while in (b), log scale was used.