Figure 7.

Effect of preprocessing on classifier performance in the KNN model. (A) Significant effect of preprocessing (artifact correction method and region of interest) on AUC scores in the k-nearest-neighbor (KNN) model. The blocking variable “Participant” is ignored in this figure to better show the effect of the preprocessing. Data cleaning was done either by combined manual and ICA artifact correction (ICA) or manually (MAN) on either motor area (MA) or whole-brain (WB) dataset. ROI, region of interest; F1, indicator of participant performance; KNN, k-nearest-neighbors; ICA, independent component analysis; MAN, manual artifact correction. ^*p < 0.05 of statistically significant contrasts as revealed by post-hoc tests (B) AUC scores across participants in the KNN model show a low amount of inter-individual variance. In this figure, the blocking variable “Participant” is accounted for. Methods indicated as follows: MA, motor area ROI; ICA, independent component analysis; WB, whole-brain ROI; MAN, manual artifact correction.