Table 3.

Classifier accuracies (ranks) for each class, the Friedman statistic (T1) and average classifier rank, for the experiments with the 4 datasets

Dataset	Classifier	Negative	Neutral	Positive (partial)	Positive (complete)	Average rank
Original T1=1.956	kNN	0 (4.5)	0.167 (5.5)	0.741 (4)	0.893 (4)	–
	LVQNN	0 (4.5)	1.0 (1)	1.0 (1)	0.929 (2)	–
	MLPI	1.0 (1)	0.667 (2)	0.778 (3)	0.893 (4)	–
	MLPII	0.5 (2)	0.5 (3)	0.889 (2)	0.893 (4)	–
	RBFNN	0 (4.5)	0.167 (5.5)	0.444 (6)	0.964 (1)	–
	PNN	0 (4.5)	0.333 (4)	0.704 (5)	0.750 (6)	–
SMOTE T1=12	kNN	0.125 (6)	0.417 (6)	0.630 (4.5)	0.893 (4)	5.125
	LVQNN	1.0 (1.5)	1.0 (1)	1.0 (1)	1.0 (1)	1.125
	MLPI	0.5 (3)	0.5 (5)	0.778 (3)	0.930 (2.5)	3.375
	MLPII	1.0 (1.5)	0.917 (2)	0.852 (2)	0.930 (2.5)	2
	RBFNN	0.25 (5)	0.667 (4)	0.444 (6)	0.393 (6)	5.25
	PNN	0.375 (4)	0.889 (3)	0.630 (4.5)	0.679 (5)	4.125
PCA T1=4.241	kNN	0 (4.5)	0.167 (5.5)	0.741 (5)	0.893 (3.5)	4.625
	LVQNN	0 (4.5)	1.0 (1)	1.0 (1)	0.929 (1.5)	2
	MLPI	0.5 (1.5)	0.5 (2.5)	0.778 (3.5)	0.929 (1.5)	2.25
	MLPII	0.5 (1.5)	0.5 (2.5)	0.889 (2)	0.893 (3.5)	2.375
	RBFNN	0 (4.5)	0.333 (4)	0.667 (6)	0.464 (6)	5.125
	PNN	0 (4.5)	0.167(5.5)	0.778 (3.5)	0.821 (5)	4.625
SMOTE+PCA T1=4.602	KNN	0.5 (3.5)	0.67 (4)	0.630 (5)	0.893 (4)	4.125
	LVQNN	1.0 (1)	1.0 (1)	0.963 (1)	1.0 (1)	1.0
	MLP	0.125 (6)	0.833 (2.5)	0.704 (4)	0.964 (2.5)	3.75
	MLPII	0.5 (3.5)	0.833 (2.5)	0.815 (2)	0.964 (2.5)	2.625
	RBFNN	0.375 (5)	0.417 (6)	0.741 (3)	0.714 (5)	4.75
	PNN	0.625 (2)	0.583 (5)	0.556 (6)	0.679 (6)	4.75