Table 3 Median performance ROC AUC scores for different AutoML models scaled according to median random forest performance. Models with the best performance for each disease are indicated in bold
From: Benchmarking AutoML frameworks for disease prediction using medical claims
Metric: ROC AUC | Lung Cancer | Prostate Cancer | Rheumatoid Arthritis | Type 2 Diabetes | IBD | CKD |
|---|---|---|---|---|---|---|
Model | Â | Â | Â | Â | Â | Â |
AutoSklearn (Average Precision) | 1.107 | 1.091 | 1.072 | 1.081 | 1.022 | 1.039 |
AutoSklearn (Balanced Accuracy) | 1.124 | 1.097 | 1.082 | 1.069 | 1.042 | 1.034 |
AutoSklearn (ROC AUC) | 1.152 | 1.109 | 1.110 | 1.091 | 1.048 | 1.041 |
H2O (AUC) | 1.159 | 1.107 | 1.107 | 1.098 | 1.042 | 1.042 |
H2O (AUCPR) | 1.159 | 1.104 | 1.106 | 1.098 | 1.042 | 1.043 |
Random Forest | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 |
TPOT (Average Precision) | 1.144 | 1.053 | 1.055 | 1.056 | 1.037 | 1.012 |
TPOT (Balanced Accuracy) | 1.071 | 1.013 | 1.058 | 1.003 | 1.032 | 1.002 |
TPOT (ROC AUC) | 1.128 | 1.103 | 1.084 | 1.075 | 1.040 | 1.013 |
Random Forest | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 | 1.000 |