Table 8.

Performance summary of classifying Car and Pedestrian events with LgR, MLP, SVM and RF classifier models using EEG and MI-based feature on the holdout test set among 738 observations where events due to pedestrian were considered as positive class. The number of observations with positive and negative class were 241 and 497 respectively. The highest accuracies obtained by using different feature sets are marked with (*).

Criteria	Using EEG-Based Features				Using MI-Based Features
	LgR	MLP	SVM	RF	LgR	MLP	SVM	RF
True Positive	17	83	186	147	75	140	207	209
False Negative	224	158	55	94	166	101	34	32
False Positive	14	62	164	5	55	27	128	12
True Negative	483	435	333	492	442	470	369	485
Sensitivity	0.07	0.34	0.77	0.61	0.31	0.58	0.86	0.87
Specificity	0.97	0.88	0.67	0.99	0.89	0.95	0.74	0.98
Precision	0.55	0.57	0.53	0.97	0.58	0.84	0.62	0.95
Recall	0.07	0.34	0.77	0.61	0.31	0.58	0.86	0.87
${\mathit{F}}_{\mathbf{1}}$ score	0.13	0.43	0.63	0.75	0.40	0.69	0.72	0.90
Accuracy	0.68	0.70	0.70	0.87 *	0.70	0.83	0.78	0.94 *
Balanced Accuracy	0.52	0.61	0.72	0.80 *	0.60	0.76	0.80	0.92 *