N | FS algorithm | FS Type | Feature set | Classifier | Performance metrics | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|
 | Accuracy | Sensitivity | Specificity | F1-score | AUC | Time to build a model (s) | |||||
1 | Without performing feature selection | NONE | Full-featured dataset | SVM | Â | 69.47 | 70.31 | 69.13 | 70.23 | 70.37 | 1635 |
95% CI | (0.71, 0.69) | (0.73, 0.68) | (0.71, 0.68) | (0.71, 0.68) | (0.71, 0.68) | ||||||
HGBÂ | Â | 62.58 | 62.72 | 61.63 | 62.18 | 62.06 | 1241 | ||||
95% CI | (0.64, 0.61) | (0.64, 0.61) | (0.61, 0.60) | (0.63, 0.61) | (0.63, 0.61) | ||||||
XGB | Â | 68.25 | 66.82 | 71.63 | 69.23 | 69.14 | 690 | ||||
95% CI | (0.69, 0.67) | (0.69, 0.67) | (0.73, 0.70) | (0.72, 0.69) | (0.71, 0.68) | ||||||
2 | Boruta-F | Wrapper-based technique | Tumor stage, tumor site, tumor size, age, metastatic status, type of treatment, lymphatic invasion, body weight | SVM | Â | 85.68 | 86.54 | 86.39 | 85.64 | 83.77 | 1419 |
95% CI | (8.401, 8.715) | (8.520, 8.795) | (8.571, 8.743) | (8.421, 8.815) | (8.274, 8.435) | ||||||
HGB | Â | 88.25 | 89.71 | 86.13 | 89.31 | 88.63 | 1360 | ||||
95% CI | (8.72, 8.947) | (8.811, 9.145) | (8.531, 8.729) | (8.80, 9.024) | (8.631, 8.985) | ||||||
XGB | Â | 82.54 | 86.43 | 87.02 | 85.97 | 86.10 | 730 | ||||
95% CI | (8.167, 8.346) | (8.517, 8.812) | (8.60, 8.827) | (8.42, 8.62) | (8.537, 8.750) | ||||||
3 | mRMR-F | Filter feature selection method | Tumor stage, history of other cancers, lymphatic invasion, tumor site, type of treatment, body weight, histological type, addiction | SVM | Â | 82.12 | 83.42 | 81.24 | 82.98 | 83.15 | 1752 |
95% CI | (8.094, 8.327) | (8.251, 8.491) | (8.02, 8.8217) | (8.147, 8.410) | (8.192, 8.551) | ||||||
HGB | Â | 81.46 | 81.42 | 81.62 | 80.52 | 80.14 | 1502 | ||||
95% CI | (8.094, 8.327) | (8.251, 8.491) | (8.02, 8.8217) | (8.147, 8.410) | (8.192, 8.551) | ||||||
XGB | Â | 80.24 | 80.52 | 80.35 | 80.26 | 81.24 | 1489 | ||||
95% CI | (7.927, 8.192) | (7.974, 8.251) | (7.914, 8.241) | (7.915, 8.15) | (8.037, 8.301) | ||||||
4 | LASSO-F | Embedded-based technique | Tumor site, tumor stage, age, type of treatment, tumor size, lymphatic invasion, weight loss, metastatic status | SVM | Â | 83.07 | 85.21 | 82.49 | 83.75 | 81.59 | 950 |
95% CI | (8.19, 8.51) | (8.420, 8.725) | (8.14, 8.397) | (8.17, 8.496) | (8.052, 8.30) | ||||||
HGB | Â | 84.12 | 84.62 | 83.19 | 82.45 | 83.09 | 1037 | ||||
95% CI | (8.274, 8.61) | (8.34, 8.61) | (8.17, 8.517) | (8.10, 8.34) | (8.21, 8.394) | ||||||
XGB | Â | 89.10 | 89.42 | 87.15 | 90.84 | 89.37 | 615 | ||||
95% CI | (8.771, 9.140) | (8.752, 9.172) | (8.682, 8.925) | (8.940, 9.153) | (8.790, 9.041) | ||||||
5 | Relief –F | Filter feature selection method | Histological type, tumor site, history of other cancers, age, vascular invasion, tumor size, type of treatment, tumor stage | SVM |  | 83.82 | 82.16 | 81.92 | 84.61 | 82.93 | 1306 |
95% CI | (8.241, 8.527) | (8.12, 8.417) | (8.034, 8.241) | (8.21, 8.516) | (8.124, 8.481) | ||||||
HGB | Â | 82.47 | 83.61 | 82.56 | 81.62 | 82.31 | 1512 | ||||
95% CI | (8.170, 8.347) | (8.21, 8.492) | (8.17, 8.397) | (8.035, 8.306) | (8.094, 8.427) | ||||||
XGB | Â | 83.75 | 84.30 | 82.07 | 83.92 | 81.01 | 1250 | ||||
95% CI | (8.201, 8.581) | (8.271, 8.609) | (8.092, 8.417) | (8.195, 8.463) | (8.037, 8.278) |