Effective hybrid feature selection using different bootstrap enhances cancers classification performance

Table 15 The proposed methods compared with the ConsistencySubsetEval method

Datasets	Train Data %	Test Data %	Over-fitting Diff. %	Pre	Rec	F1-score	NO.F	F-Time (sec)	C-Time (sec)	AUC	Var.	ACC %
J48 classifier
RNA gene	93.106	91.136	1.97	0.880	0.869	0.869	3	1.850	0.001	0.963	0.000641	91.136
DNA CNV	64.842	63.921	0.921	0.550	0.549	0.543	42	1.600	0.012	0.816	0.001172	63.921
Parkinson’s disease	86.346	80.279	6.067	0.765	0.688	0.707	11	1.100	0.003	0.738	0.002407	80.279
Dermatology diseases	87.918	87.740	0.178	0.751	0.755	0.742	12	0.102	0.0006	0.946	0.002407	87.740
BreastEW	96.993	94.380	2.613	0.950	0.933	0.939	8	0.090	0.0008	0.971	0.000873	94.380
Naïve base classifier
RNA gene	97.545	97.380	0.165	0.972	0.970	0.970	3	1.850	0.001	0.994	0.000188	97.380
DNA CNV	73.015	71.606	1.409	0.682	0.698	0.680	42	1.600	0.006	0.923	0.001276	71.606
Parkinson’s disease	76.940	75.516	1.424	0.666	0.599	0.605	11	1.100	0.0005	0.729	0.003051	75.516
Dermatology diseases	90.265	89.839	0.426	0.878	0.901	0.867	12	0.102	0.0007	0.995	0.002740	89.839
BreastEW	94.630	94.201	0.429	0.943	0.934	0.937	8	0.090	0.002	0.988	0.000686	94.201
K-nearest neighbors(KNN) classifier
RNA gene	97.517	97.131	0.386	0.963	0.964	0.962	3	1.850	0.001	0.993	0.000311	97.131
DNA CNV	82.735	77.059	5.676	0.750	0.680	0.685	42	1.600	0.0001	0.888	0.000465	77.059
Parkinson’s disease	79.100	69.698	9.402	0.557	0.524	0.517	11	1.100	0.002	0.583	0.002962	69.698
Dermatology diseases	97.632	95.916	1.716	0.962	0.943	0.947	12	0.102	0.0008	0.994	0.001339	95.916
BreastEW	95.704	93.496	2.208	0.937	0.925	0.929	8	0.090	0.0009	0.965	0.000621	93.496

ISSN: 1756-0381