TABLE 3.
Estimation of the accuracy of the gene expression test for prostate cancer data under different conditions of gene selection and classification
Data dividing | No. Layers | Feature selection by wrapper method | Feature selection by ensemble model | ||||
---|---|---|---|---|---|---|---|
Max | Mean | Min | Max | Mean | Min | ||
10‐fold (1) | Low | 0.89 ± (0.046) | 0.86 ± (0.071) | 0.85 ± (0.141) | 0.91 ± (0.032) | 0.88 ± (0.048) | 0.86 ± (0.132) |
Med | 0.94 ± (0.033) | 0.90 ± (0.060) | 0.88 ± (0.073) | 0.98 ± (0.026) | 0.96 ± (0.031) | 0.93 ± (0.056) | |
High | 0.90 ± (0.057) | 0.86 ± (0.058) | 0.85 ± (0.129) | 0.94 ± (0.044) | 0.91 ± (0.058) | 0.89 ± (0.108) | |
10‐fold (2) | Low | 0.90 ± (0.068) | 0.86 ± (0.092) | 0.83 ± (0.121) | 0.93 ± (0.045) | 0.90 ± (0.063) | 0.87 ± (0.119) |
Med | 0.94 ± (0.053) | 0.92 ± (0.064) | 0.90 ± (0.070) | 100 | 0.97± (0.024) | 0.94 ± (0.051) | |
High | 0.91 ± (0.073) | 0.89 ± (0.084) | 0.87 ± (0.113) | 0.95 ± (0.034) | 0.93 ± (0.055) | 0.90 ± (0.095) | |
10‐fold (3) | Low | 0.90 ± (0.056) | 0.88 ± (0.094) | 0.86 ± (0.118) | 0.92 ± (0.035) | 0.90 ± (0.044) | 0.87 ± (0.093) |
Med | 0.94 ± (0.034) | 0.90 ± (0.056) | 0.88 ± (0.096) | 0.98 ± (0.025) | 0.95 ± (0.035) | 0.92 ± (0.048) | |
High | 0.91 ± (0.076) | 0.89 ± (0.073) | 0.87 ± (0.108) | 0.94 ± (0.042) | 0.91 ± (0.057) | 0.88 ± (0.084) | |
10‐fold (4) | Low | 0.89 ± (0.067) | 0.87 ± (0.075) | 0.86 ± (0.123) | 0.93 ± (0.035) | 0.91 ± (0.067) | 0.87 ± (0.083) |
Med | 0.95 ± (0.048) | 0.93 ± (0.041) | 0.91 ± (0.087) | 100 | 0.97 ± (0.024) | 0.94 ± (0.041) | |
High | 0.89 ± (0.074) | 0.87 ± (0.085) | 0.86 ± (0.106) | 0.94 ± (0.041) | 0.91 ± (0.051) | 0.89 ± (0.072) | |
10‐fold (5) | Low | 0.89 ± (0.056) | 0.85 ± (0.074) | 0.84 ± (0.143) | 0.92 ± (0.033) | 0.88 ± (0.061) | 0.85 ± (0.128) |
Med | 0.94 ± (0.041) | 0.92 ± (0.060) | 0.90 ± (0.082) | 0.98 ± (0.020) | 0.95 ± (0.049) | 0.91 ± (0.086) | |
High | 0.91 ± (0.050) | 0.89 ± (0.071) | 0.88 ± (0.113) | 0.96 ± (0.039) | 0.93 ± (0.064) | 0.90 ± (0.096) |