Figure 5.

Classification analysis shows functional connectivity outperforms regional power in classification modelling at the delta frequency. The PCA results show that the functional connectivity data successfully separated the two participant groups with ML selected features, while the delta power data failed to achieve so. The permutation tests for both SVM and PLS-DA modelling showed that the delta functional connectivity models are more significant (smaller P-values) than the regional power models. When tested on the entire subject set (training data), the delta power model failed to achieve the expected optimal performance in AUC, whereas the functional connectivity counterpart achieved such.