Table 1 Performance of five feature selection strategies for identifying placental cell subpopulations on four machine learning models (Independent dataset)
Base classifier | Feature selection | Feature numbers | Accuracy (%) | Precision (%) | Recall (%) | F1 measure (%) |
|---|---|---|---|---|---|---|
KNN | PCA | 160 | 71.67 | 79.69 | 68.57 | 71.32 |
RFC | PCA | 3000 | 91.13 | 93.43 | 90.80 | 91.87 |
SVM | PCA | 1200 | 90.12 | 91.93 | 89.98 | 90.87 |
XGBoost | PCA | 2000 | 92.11 | 93.03 | 91.72 | 92.32 |
KNN | MIC | 210 | 88.39 | 90.43 | 88.40 | 88.83 |
RFC | MIC | 260 | 92.40 | 93.63 | 92.28 | 92.83 |
SVM | MIC | 160 | 92.65 | 93.22 | 92.89 | 93.14 |
XGBoost | MIC | 310 | 93.07 | 93.76 | 92.87 | 93.25 |
KNN | TURF | 110 | 87.88 | 90.61 | 86.80 | 87.85 |
RFC | TURF | 310 | 92.35 | 93.90 | 92.24 | 92.86 |
SVM | TURF | 210 | 92.31 | 93.23 | 92.48 | 92.75 |
XGBoost | TURF | 110 | 92.61 | 92.98 | 92.46 | 92.65 |
KNN | F-score | 310 | 85.43 | 88.74 | 85.71 | 85.97 |
RFC | F-score | 610 | 92.02 | 93.43 | 92.03 | 92.62 |
SVM | F-score | 710 | 92.02 | 92.87 | 92.37 | 92.54 |
XGBoost | F-score | 410 | 92.10 | 93.04 | 92.31 | 92.64 |
KNN | ANOVA | 360 | 84.00 | 87.28 | 83.99 | 83.91 |
RFC | ANOVA | 810 | 92.10 | 93.75 | 91.95 | 92.65 |
SVM | ANOVA | 710 | 92.02 | 93.33 | 92.51 | 92.83 |
XGBoost | ANOVA | 460 | 92.11 | 93.24 | 92.22 | 92.67 |