“Dimension Reduction: Feature Subset” Method for Selecting the Best Index Combination in Reputation Evaluation of Crowdsourcing Participants

Table 4

Accuracy of ten-fold cross-validation of feature subset selected by classifier.

Feature subset	ReliefF-DT	ReliefF-SVM	ReliefF-BPNN	ReliefF-RBFNN	ReliefF-NB	ReliefF-KNN
5	0.9036	0.9104	0.9050	0.8930	0.8592	0.8835
6	0.9043	0.9131	0.9087	0.8967	0.8643	0.8826
7	0.9026	0.9112	0.9085	0.8974	0.8665	0.8875
8	0.9031	0.9132	0.9088	0.8981	0.8788	0.8958
9	0.9055	0.9147	0.9086	0.8987	0.8843	0.8934
10	0.9012	0.9117	0.9099	0.8938	0.8874	0.8897
11	0.8993	0.9120	0.9126	0.8920	0.8869	0.8905
12	0.8998	0.9111	0.9132	0.8860	0.8856	0.8894
13	0.9034	0.9127	0.9108	0.8850	0.8890	0.8895
14	0.9032	0.9140	0.9109	0.8800	0.8818	0.8872
15	0.9012	0.9104	0.9100	0.8740	0.8832	0.8870
16	0.8997	0.9112	0.9130	0.8710	0.8791	0.8926
17	0.8987	0.9137	0.9100	0.8670	0.8760	0.8900
18	0.9000	0.9089	0.9110	0.8630	0.8781	0.8870
19	0.8977	0.9052	0.9130	0.8570	0.8782	0.8848
20	0.8989	0.9010	0.9100	0.8560	0.8781	0.8830

The best result for each classifier is in bold.