Table 4.
Comparison of accuracy and MAUC among five classification methods, based on 1000 runs of k-fold (k = 10) cross-validation.
| METHOD | Accuracy (MEAN ± SD) MAUC (95% CI) | Overall Accuracy/Overall MAUC | ||||
|---|---|---|---|---|---|---|
| HS | CS | HR | CR | |||
| LDA | 97.0% ± 5.3% 0.982~0.986 |
98.9% ± 3.2% 0.993~0.995 |
94.3% ± 6.7% 0.967~0.971 |
94.7% ± 6.6% 0.969~0.974 |
96.2%/0.980 | |
| PLS-DA | 68.4% ± 16.0% 0.835~0.845 |
56.1% ± 15.2% 0.776~0.785 |
56.5% ± 18.4% 0.778~0.790 |
50.0% ± 22.4% 0.748~0.762 |
57.8%/0.794 | |
| KNN | K = 1 | 98.3% ± 4.3% 0.990~0.993 |
98.9% ± 3.2% 0.993~0.995 |
93.9% ± 7.0% 0.966~0.970 |
93.7% ± 7.2% 0.965~0.970 |
96.3%/0.981 |
| K = 2 | 97.4% ± 5.0% 0.984~0.988 |
98.2% ± 4.2% 0.988~0.991 |
94.0% ± 7.1% 0.964~0.969 |
94.2% ± 7.2% 0.964~0.969 |
96.0%/0.977 | |
| K = 3 | 97.1% ± 5.6% 0.984~0.988 |
98.7% ± 3.4% 0.993~0.994 |
92.4% ± 7.5% 0.957~0.962 |
93.2% ± 7.4% 0.963~0.968 |
95.4%/0.977 | |
| K = 4 | 96.1% ± 6.0% 0.977~0.981 |
98.2% ± 4.1% 0.989~0.992 |
93.2% ± 7.3% 0.960~0.966 |
94.3% ± 7.1% 0.966~0.971 |
95.5%/0.975 | |
| K = 5 | 95.7% ± 6.6% 0.976~0.980 |
98.6% ± 3.6% 0.991~0.994 |
93.2% ± 7.6% 0.961~0.967 |
93.2% ± 8.0% 0.962~0.967 |
95.2%/0.975 | |
| K = 6 | 95.1% ± 6.7% 0.971~0.976 |
98.3% ± 4.0% 0.989~0.992 |
93.2% ± 7.5% 0.961~0.967 |
93.2% ± 7.9% 0.961~0.967 |
95.0%/0.973 | |
| SVM | linear | 94.1% ± 7.8% 0.966~0.970 |
97.8% ± 4.7% 0.986~0.990 |
91.4% ± 8.6% 0.952~0.958 |
97.0% ± 5.0% 0.982~0.986 |
95.1%/0.974 |
| polynomial | 93.6% ± 8.3% 0.964~0.969 |
97.1% ± 5.3% 0.983~0.987 |
68.2% ± 13.5% 0.832~0.841 |
97.9% ± 4.4% 0.987~0.990 |
89.2%/0.944 | |
| RBF | 92.6% ± 8.6% 0.959~0.965 |
95.8% ± 6.2% 0.976~0.980 |
89.1% ± 9.9% 0.940~0.947 |
91.0% ± 9.4% 0.952~0.957 |
92.1%/0.960 | |
| sigmoid | 34.7% ± 15.0% 0.662~0.671 |
8.1% ± 12.8% 0.497~0.506 |
42.8% ± 16.0% 0.693~0.703 |
14.3% ± 11.8% 0.568~0.577 |
25.0%/0.610 | |
| RF | 91.44% ± 9.3% 0.953~0.959 |
93.8% ± 7.2% 0.966~0.970 |
84.1% ± 11.1% 0.913~0.920 |
83.3% ± 11.7% 0.909~0.917 |
88.1%/0.939 | |