Table 1 The performance of predicting content polarity using different input features and their combinations, both before and after adding EEG data. Adding EEG data significantly improved ACC, F1, and AUC for all feature combinations. When using the combination of EEG, Behavior, Emotion, and Sentiment features, the model achieved an ACC of 0.646, an F1 score of 0.644, and an AUC of 0.685.

Input Features				ACC	F1	AUC
EEG	Like	Emotion	Sentiment	Mean ± SEM	Mean ± SEM	Mean ± SEM
-	-	-	\(\checkmark\)	\(0.592^* \pm 0.029\)	\(0.588^* \pm 0.028\)	\(0.662 \pm 0.038\)
-	-	\(\checkmark\)	-	\(0.517^* \pm 0.041\)	\(0.517^* \pm 0.040\)	\(0.504^* \pm 0.033\)
-	-	\(\checkmark\)	\(\checkmark\)	\(0.497^* \pm 0.030\)	\(0.497^* \pm 0.029\)	\(0.530^* \pm 0.026\)
-	\(\checkmark\)	-	-	\(0.557^* \pm 0.015\)	\(0.485^* \pm 0.021\)	\(0.548^* \pm 0.023\)
-	\(\checkmark\)	-	\(\checkmark\)	\(0.607 \pm 0.023\)	\(0.598 \pm 0.023\)	\(0.630 \pm 0.034\)
-	\(\checkmark\)	\(\checkmark\)	-	\(0.552^* \pm 0.028\)	\(0.550^* \pm 0.027\)	\(0.514^* \pm 0.032\)
-	\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	\(0.537^* \pm 0.014\)	\(0.536^* \pm 0.014\)	\(0.581^* \pm 0.025\)
\(\checkmark\)	-	-	-	\(0.641 \pm 0.024\)	\(0.639 \pm 0.024\)	\(0.672 \pm 0.029\)
\(\checkmark\)	-	-	\(\checkmark\)	\(0.622 \pm 0.019\)	\(0.620 \pm 0.019\)	\(0.677 \pm 0.016\)
\(\checkmark\)	-	\(\checkmark\)	-	\(0.597^* \pm 0.017\)	\(0.594^* \pm 0.017\)	\(0.639 \pm 0.025\)
\(\checkmark\)	-	\(\checkmark\)	\(\checkmark\)	\(0.631 \pm 0.025\)	\(0.629 \pm 0.024\)	\(0.672 \pm 0.019\)
\(\checkmark\)	\(\checkmark\)	-	-	\(0.627 \pm 0.022\)	\(0.623 \pm 0.023\)	\(0.670 \pm 0.030\)
\(\checkmark\)	\(\checkmark\)	-	\(\checkmark\)	0.647 ± 0.019	0.644 ± 0.020	0.687 ± 0.021
\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	-	\(0.577^* \pm 0.026\)	\(0.572^* \pm 0.024\)	\(0.634 \pm 0.024\)
\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	\(\checkmark\)	\(0.646 \pm 0.016\)	0.644 ± 0.016	\(0.685 \pm 0.027\)

Quick links

Search