Table 7 Comparison of patient similarity model performance with other models
From: Patient similarity analytics for explainable clinical risk prediction
Model | AUROC (95% CI) |
|---|---|
Patient similarity (K = 10)—weighted | 0.718 (0.697 to 0.739) |
Patient similarity (K = 10)—unweighted | 0.688 (0.667 to 0.709) |
Logistic regression | 0.695 (0.672 to 0.718) |
Random forest | 0.764 (0.744 to 0.784) |
Support vector machine (kernel = linear) | 0.766 (0.746 to 0.785) |