Automatic detection of attachment style in married couples through conversation analysis

EURASIP Journal on Audio, Speech, and Music Processing

Table 7 Accuracy, precision, recall and F1 scores with i-vectors and low-level features are shown for the investigated classifiers and feature selection algorithms. Female and male features are used together and separately

	Algorithm	Gender	Accuracy	Precision	Recall	F1 score
SVM	mRMR(30)	Male	0.68	0.66	0.72	0.69
	JMI (50)	Female	0.66	0.67	0.63	0.65
	mRMR (15)	Both	0.75	0.76	0.72	0.74
Decision tree	mRMR(10)	Male	0.69	0.68	0.68	0.68
	JMI(15)	Female	0.68	0.69	0.65	0.67
	MIM (5)	Both	0.68	0.68	0.67	0.68
Random forest	JMI(10)	Male	0.77	0.79	0.72	0.75
	JMI(10)	Female	0.77	0.79	0.72	0.75
	JMI (10)	Both	0.77	0.8	0.72	0.76
AdaBoost	JMI(10)	Male	0.72	0.71	0.70	0.71
	JMI(30)	Female	0.74	0.73	0.75	0.74
	MIM (30)	Both	0.79	0.79	0.77	0.78
Gradient boosting	JMI(10)	Male	0.74	0.75	0.7	0.73
	JMI(15)	Female	0.75	0.76	0.72	0.74
	JMI (10)	Both	0.76	0.76	0.76	0.76
Extra tree	JMI(5)	Male	0.77	0.83	0.69	0.75
	JMI(15)	Female	0.77	0.76	0.77	0.77
	JMI (50)	Both	0.79	0.79	0.77	0.78
XGBoost	JMI(10)	Male	0.76	0.75	0.77	0.76
	mRMR(15)	Female	0.76	0.77	0.74	0.75
	JMI (10)	Both	0.84	0.76	0.79	0.78
Artificial neural network	mRMR(50)	Male	0.68	0.67	0.69	0.68
	JMI(50)	Female	0.71	0.71	0.69	0.7
	MIM (30)	Both	0.78	0.76	0.79	0.78