Effective Evolutionary Multilabel Feature Selection under a Budget Constraint

Table 3

Comparison results for multilabel feature selection methods in terms of multilabel accuracy and Hamming loss (mean std. deviation). The ✓ symbol indicates the method that achieves the best performance for each dataset.

Methods	Evaluation measure
	Multi-label accuracy		Hamming loss
	Proposed	RGA	Proposed	RGA
Birds	0.497 0.048✓	0.459 0.048	0.055 0.005✓	0.056 0.004
Emotions	0.460 0.020✓	0.447 0.029	0.243 0.022✓	0.252 0.016
Enron	0.360 0.021✓	0.271 0.042	0.056 0.002✓	0.060 0.001
Genbase	0.886 0.041✓	0.155 0.097	0.011 0.004✓	0.042 0.003
LLog	0.213 0.027✓	0.166 0.026	0.016 0.001	0.016 0.001✓
Mediamill	0.366 0.002✓	0.359 0.005	0.034 0.000	0.034 0.000✓
Medical	0.517 0.048✓	0.097 0.046	0.018 0.002✓	0.026 0.002
Scene	0.408 0.019✓	0.352 0.030	0.157 0.002	0.154 0.005✓
Slashdot	0.144 0.017✓	0.031 0.011	0.048 0.001✓	0.053 0.000
TMC2007	0.372 0.005✓	0.318 0.020	0.084 0.001✓	0.088 0.001
Yeast	0.465 0.013✓	0.442 0.019	0.224 0.006✓	0.225 0.010
Arts	0.140 0.012✓	0.049 0.015	0.060 0.001✓	0.063 0.001
Business	0.678 0.011	0.678 0.008✓	0.029 0.001	0.029 0.001✓
Education	0.109 0.019✓	0.033 0.011	0.042 0.001✓	0.044 0.001
Entertain	0.233 0.016✓	0.128 0.042	0.058 0.000✓	0.065 0.002
Health	0.510 0.018✓	0.402 0.016	0.040 0.001✓	0.049 0.001
Reference	0.382 0.044	0.393 0.011✓	0.030 0.001✓	0.034 0.001
Science	0.120 0.011✓	0.042 0.015	0.034 0.001✓	0.036 0.001
Social	0.546 0.018✓	0.134 0.060	0.024 0.001✓	0.030 0.001
Society	0.304 0.135✓	0.280 0.146	0.055 0.001✓	0.059 0.001