Self-Trained LMT for Semisupervised Learning

Table 5

Average rankings of the algorithms in 10% labeled ratio (Friedman) and Holm/Hochberg (alpha = 0.05).

Algorithm	Friedman; ranking	value	Holm/; Hochberg test
Self-LMT	6.8750	—	—
Tri-training (C45)	8.7019	1.5141E − 01	0.0500
Cobagging (C45)	9.2019	6.7671E − 02	0.0250
Co-Forest	9.6346	3.0238E − 02	0.0167
Democratic-Co	9.8558	1.9252E − 02	0.0125
Self-training (C45)	10.2308	8.4117E − 03	0.0100
Cotraining (SMO)	11.2115	6.6111E − 04	0.0083
SETRED	11.3846	3.9842E − 04	0.0071
Tri-training (SMO)	11.5000	2.8153E − 04	0.0063
SNNRCE	11.5000	2.8153E − 04	0.0056
Cotraining (C45)	11.5096	2.7340E − 04	0.0050
Cobagging (SMO)	11.5865	2.1587E − 04	0.0045
DE-tri-training (C45)	11.7788	1.1778E − 04	0.0042
Tri-training (NN)	11.9135	7.6088E − 05	0.0038
ADE-Co-Forest	12.0481	4.8633E − 05	0.0036
DE-tri-training (SMO)	12.1827	3.0754E − 05	0.0033
Self-training (SMO)	12.3173	1.9242E − 05	0.0031
Self-training (NN)	12.6827	5.1049E − 06	0.0029
APSSC	13.5769	1.4202E − 07	0.0028
CLCC	13.7500	6.7193E − 08	0.0026
Rel-Rasco (NB)	14.4231	3.0843E − 09	0.0025
Rasco (C45)	15.1346	8.8277E − 11	0.0024