VitalDB - AnesLLM: a dataset and benchmark based on real intraoperative anaesthetic decisions 1.0
| strategy | method |
| knn | kNN Euclidean K=50 (baseline) |
| random_forest | Random Forest, probability out-of-fold (3 folds) |
| lightgbm | LightGBM, probabilities out-of-fold (3 folds) |
| logistic | Softmax logistic regression, out-of-fold (3 folds) |
| centroid | affinity = softmax(distance to class centroid) |
| dwknn | kNN weighted by 1/distance |
| umap_knn | UMAP (10 dims) + kNN K=50 |