Feature Selection Using Maximum Feature Tree Embedded with Mutual Information and Coefficient of Variation for Bird Sound Classification

Table 4

Comparison of Kappa, accuracy, and F₁ scores with feature selection methods in Birds dataset.

Evaluation indicator	Classifier	Feature selection method
		Cor	GR	IG	OR	RF	SU	CS	MICV
		Number of features|the highest value
Kappa	J48	72 | 0.85	60 | 0.85	59 | 0.86	57 | 0.86	73 | 0.83	68 | 0.84	51 | 0.84	38 | 0.88
	NB	68 | 0.79	37 | 0.86	36 | 0.87	39 | 0.86	70 | 0.77	36 | 0.87	58 | 0.79	46 | 0.88
	SVM	63 | 0.93	44 | 0.93	52 | 0.93	61 | 0.93	64 | 0.95	46 | 0.93	65 | 0.95	51 | 0.95
	RFs	71 | 0.97	50 | 0.97	36 | 0.97	73 | 0.97	72 | 0.95	70 | 0.97	53 | 0.96	30 | 0.97
Accuracy	J48	72 | 88.18	60 | 88.18	59 | 89.09	57 | 89.09	73 | 86.36	65 | 87.27	52 | 87.27	36 | 90.96
	NB	64 | 83.63	37 | 89.09	36 | 88.18	39 | 89.09	70 | 81.81	36 | 90.00	58 | 83.63	34 | 90.90
	SVM	63 | 94.54	44 | 94.45	41 | 94.00	61 | 94.45	64 | 96.36	46 | 94.45	65 | 96.36	51 | 93.63
	RFs	71 | 98.12	50 | 98.12	36 | 98.12	73 | 98.18	72 | 96.36	70 | 98.12	53 | 97.27	30 | 98.12
F1 score	J48	72 | 0.88	60 | 0.88	59 | 0.89	57 | 0.89	73 | 0.86	65 | 0.87	51 | 0.87	38 | 0.87
	NB	64 | 0.83	37 | 0.89	36 | 0.89	39 | 0.89	70 | 0.81	37 | 0.90	58 | 0.83	34 | 0.90
	SVM	63 | 0.94	49 | 0.94	41 | 0.94	61 | 0.94	64 | 0.96	46 | 0.94	65 | 0.96	51 | 0.93
	RFs	71 | 0.96	50 | 0.98	55 | 0.98	73 | 0.98	73 | 0.96	70 | 0.98	54 | 0.97	30 | 0.98