Table 4.
Average performance of logistic regression and machine learning models for predicting prognostic biomarkers and molecular subtypes.
| Lymph node | Tumor grade | Tumor size | ER | PR | HER2 | Ki67 | Molecular subtype | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Accuracy | AUC | Accuracy | AUC | Accuracy | AUC | Accuracy | AUC | Accuracy | AUC | Accuracy | AUC | Accuracy | AUC | Accuracy | AUC | |
| Logistic regression | 62% | 0.66 | 67% | 0.71 | 64% | 0.69 | 70% | 0.73 | 66% | 0.68 | 78% | 0.69 | 65% | 0.70 | 48% | 0.69 |
| Decision tree | 66% | 0.65 | 71% | 0.69 | 65% | 0.65 | 72% | 0.69 | 69% | 0.67 | 77% | 0.67 | 66% | 0.66 | 50% | 0.63 |
| Naïve Bayes | 53% | 0.58 | 63% | 0.70 | 58% | 0.63 | 68% | 0.71 | 65% | 0.65 | 72% | 0.69 | 59% | 0.69 | 49% | 0.70 |
| Random forest | 78% | 0.86 | 80% | 0.88 | 77% | 0.85 | 82% | 0.88 | 78% | 0.85 | 83% | 0.88 | 77% | 0.85 | 66% | 0.82 |
| SVM | 57% | 0.34 | 63% | 0.35 | 48% | 0.42 | 67% | 0.44 | 64% | 0.39 | 79% | 0.47 | 53% | 0.35 | 41% | 0.65 |
| ANN | 64% | 0.68 | 68% | 0.73 | 65% | 0.71 | 75% | 0.77 | 69% | 0.72 | 76% | 0.73 | 66% | 0.71 | 35% | 0.72 |
ER estrogen receptor, PR progesterone receptor, HER2 human epidermal growth factor receptor 2, AUC the area under the receiver-operating-characteristic curve, SVM support vector machine, ANN artificial neural network.