Self-Trained LMT for Semisupervised Learning

Table 8

Average rankings of the algorithms in 40% labeled ratio (Friedman) and Holm/Hochberg (alpha = 0.05).

Algorithm	Friedman; ranking	value	Holm/; Hochberg test
Self-LMT	5.0865	—	—
Democratic-Co	9.1923	1.2641E − 03	0.0500
Tri-training (C45)	9.2115	1.1990E − 03	0.0250
Cotraining (SMO)	9.4135	6.7962E − 04	0.0167
Cotraining (C45)	9.8942	1.5989E − 04	0.0125
Self-training (C45)	9.9519	1.3319E − 04	0.0100
Co-Forest	10.0577	9.4794E − 05	0.0083
Cobagging (SMO)	10.7212	9.6656E − 06	0.0071
Self-training (SMO)	10.7308	9.3332E − 06	0.0063
Tri-training (SMO)	10.8942	5.1049E − 06	0.0056
Cobagging (C45)	11.0096	3.3028E − 06	0.0050
DE-tri-training (SMO)	11.6923	2.1358E − 07	0.0045
DE-tri-training (C45)	12.1827	2.5157E − 08	0.0042
ADE-Co-Forest	12.2596	1.7753E − 08	0.0038
Rasco (C45)	12.2692	1.6992E − 08	0.0036
SETRED	12.3269	1.3048E − 08	0.0033
Self-training (NN)	13.0769	3.5107E − 10	0.0031
SNNRCE	13.5385	3.2060E − 11	0.0029
Tri-training (NN)	14.1827	9.1543E − 13	0.0028
APSSC	14.2404	6.5767E − 13	0.0026
Rel-Rasco (NB)	14.4808	1.6224E − 13	0.0025
CLCC	16.5865	1.7128E − 19	0.0024