Table 3.
Diagnostic performance of classification models.
| Models | Training group | Validation group | |||||||
|---|---|---|---|---|---|---|---|---|---|
| WHO Grade I | WHO Grade II | WHO Grade III | Kappa value | WHO Grade I | WHO Grade II | WHO Grade III | Kappa value | ||
| LDA | Distance Correlation | 0.928 | 0.865 | 0.882 | 0.578 (Accuracy = 75.4%) |
0.884 | 0.820 | 0.846 | 0.563 (Accuracy = 73.0%) |
| LASSO | 0.955 | 0.914 | 0.915 | 0.693 (Accuracy = 80.8%) |
0.934 | 0.846 | 0.783 | 0.603 (Accuracy = 75.6%) |
|
| GBDT | 0.928 | 0.950 | 0.908 | 0.570 (Accuracy = 73%) |
0.886 | 0.854 | 0.887 | 0.572 (Accuracy = 73.3%) |
|
| SVM | Distance Correlation | 0.870 | 0.831 | 0.876 | 0.356 (Accuracy = 61.1%) |
0.845 | 0.798 | 0.845 | 0.274 (Accuracy = 57.6%) |
| LASSO | 0.898 | 0.806 | 0.877 | 0.373 (Accuracy = 62.0%) |
0.840 | 0.772 | 0.833 | 0.298 (Accuracy = 59.0%) |
|
| GBDT (Overfitting) | – | – | – | – | – | – | – | ||
LDA, Linear Discriminate Analysis; SVM, Support Vector Machine; LASSO, Least absolute shrinkage and selection operator; GBDT, gradient boosting decision tree.