Research Article

A Parallel Genetic Algorithm Based Feature Selection and Parameter Optimization for Support Vector Machine

Table 3

Comparisons between CGPGA-SVM and grid search.

Dataset⁢CGPGA-SVM⁢Grid searchp values
Tenfold cross validation accuracy (%)Number of selected featuresTenfold cross validation accuracy (%)Number of selected features

Ionosphere 98.85 ± 2.0113.8 ± 3.4693.16 ± 3.0034.0 ± 0.000.005
Breast cancer98.53 ± 1.162.60 ± 0.8496.77 ± 1.8110.0 ± 0.000.039
Australia90.13 ± 2.395.10 ± 1.6085.51 ± 4.4614.0 ± 0.000.011
Diabetes81.76 ± 3.373.90 ± 0.9976.44 ± 3.748.00 ± 0.000.014
Vehicle86.05 ± 3.5410.3 ± 1.3478.60 ± 3.2918.0 ± 0.000.005
Vowel99.29 ± 0.686.70 ± 0.4898.88 ± 1.0013.0 ± 0.000.037
Car99.83 ± 0.396.00 ± 0.0099.48 ± 0.516.00 ± 0.000.046
Splice92.66 ± 1.4326.7 ± 3.5392.03 ± 1.5260.0 ± 0.000.386
DNA96.79 ± 1.3187.1 ± 4.5396.05 ± 1.09 180. ± 0.000.005
WaveForm87.81 ± 1.6021.1 ± 2.6986.68 ± 1.4740.0 ± 0.000.037
Svmguide196.66 ± 0.762.40 ± 0.5296.32 ± 0.714.00 ± 0.000.594
Mushrooms100.0 ± 0.0042.5 ± 3.5499.98 ± 0.04 112. ± 0.000.317

indicates significance at 0.005 level.