Research Article

Evolutionary Multilabel Feature Selection Using Promising Feature Subset Generation

Table 3

Comparison results in terms of hamming loss.

MethodBirdsEnronLlogMediamillMedical
Proposed0.061 ± 0.0080.060 ± 0.0030.016 ± 0.0010.034 ± 0.0000.020 ± 0.001
GA0.072 ± 0.0150.100 ± 0.0330.075 ± 0.0730.048 ± 0.0220.023 ± 0.001
NSGA0.064 ± 0.0080.104 ± 0.0260.072 ± 0.0710.054 ± 0.0320.022 ± 0.001
MPSOFS0.135 ± 0.0180.198 ± 0.0070.292 ± 0.0080.174 ± 0.0050.023 ± 0.001

MethodTMC2007BusinessEducationEntertainmentHealth
Proposed0.088 ± 0.0020.029 ± 0.0010.042 ± 0.0010.055 ± 0.0010.039 ± 0.001
GA0.088 ± 0.0040.035 ± 0.0050.046 ± 0.0030.070 ± 0.0070.051 ± 0.004
NSGA0.086 ± 0.0040.037 ± 0.0100.048 ± 0.0040.068 ± 0.0040.054 ± 0.005
MPSOFS0.117 ± 0.0020.079 ± 0.0030.061 ± 0.0020.105 ± 0.0030.067 ± 0.002

MethodReferenceScienceSocialSocietyAvg. rank
Proposed0.034 ± 0.0060.035 ± 0.0030.025 ± 0.0030.054 ± 0.0021.07
GA0.055 ± 0.0130.051 ± 0.0120.042 ± 0.0070.062 ± 0.0052.71
NSGA0.047 ± 0.0080.045 ± 0.0080.040 ± 0.0100.060 ± 0.0022.29
MPSOFS0.086 ± 0.0050.110 ± 0.0050.070 ± 0.0020.144 ± 0.0073.93