Table 11.

Results of proposed approach on DAIC Dataset

S.no.	ML Classifiers	Fused	All features		Pearson correlation		PCA
		modalities	features		reduced feature vector
			#	Acc	#	Acc	#	Acc
1	Logistic Regression	Audio	82	70	50	81	20-25	68
2		Video	230	81	72	83	20-25	80
3		Video + Audio	312	83	122	86	30-35	83
4	Decision Tree	Audio	82	62	50	71	20-25	62
5		Video	230	80	72	80	20-25	82
6		Video + Audio	312	80	122	82	30-35	82
7	Naive Bayes	Audio	82	55	50	70	20-25	70
8		Video	230	80	72	75	20-25	80
9		Video + Audio	312	80	122	82	30-35	80
10	Random Forest	Audio	82	74	50	66	20-25	64
11		Video	230	85	72	80	20-25	81
12		Video + Audio	312	85	122	85	30-35	81
13	Support Vector Machines	Audio	82	74	50	68	20-25	67
14		Video	230	85	72	83	20-25	82
15		Video + Audio	312	85	122	86	30-35	83

# - number of features in the feature vector. Acc-Accuracy, and BOLD: Best accuracies obtained Note- DAIC dataset does not contain all the low-level openface feature sets. Hence we extracted statistical feature vector on the available low-level feature vector of DAIC dataset