Figure 7: Mutual Information analysis for understanding feature significance in binary and multi-class classification.

(A) Linear Trend analysis on Mutual Information Scores of 30 binned discrete features with the Binary-Class target, and (B) Linear Trend Analysis on Mutual Information Scores of 30 binned discrete features with the Multi-Class target. Each point indicates the mutual information score for a binned discrete input feature with target, and a fitted trend line showing the overall trend. The downward trends in both plots highlight that the significant dependence of the target on these features decreases over time, with the dependency diminishing as time approaches the end of the stimulation period (3000ms). This analysis aids in understanding which features are most influential in predicting the target in both binary and multi-class scenarios.