Table 5.
Model performance for supratherapeutic.
| Models | Precision, % | Recall, % | F1 score, % | |
| Data set 1: MIMIC-IIIa (intravenous push patients) | ||||
| Random forest | 75.95 | 75.00 | 75.47 | |
| Adaptive boosting | 67.37 | 80.00 | 73.14 | |
| Support vector machine | 100 | 57.50 | 73.02 | |
| Extreme gradient boosting | 80.77 | 78.75 | 79.75 | |
| Shallow neural network | 88.00 | 82.50 | 85.16 | |
| Data set 2: MIMIC-III (intravenous drip patients) |
|
|
||
| Random forest | 67.50 | 54.00 | 60.00 | |
| Adaptive boosting | 75.00 | 72.00 | 73.47 | |
| Support vector machine | 100 | 58.00 | 73.42 | |
| Extreme gradient boosting | 76.47 | 78.00 | 77.23 | |
| Shallow neural network | 86.54 | 90.00 | 88.24 | |
| Data set 3: eICUb (intravenous drip patients) |
|
|
||
| Random forest | 63.64 | 87.50 | 73.68 | |
| Adaptive boosting | 75.00 | 87.50 | 80.77 | |
| Support vector machine | 95.24 | 83.33 | 88.89 | |
| Extreme gradient boosting | 81.48 | 91.67 | 86.27 | |
| Shallow neural network | 95.45 | 87.50 | 91.30 | |
aMultiparameter Intelligent Monitoring In Intensive Care III database.
beICU: e–Intensive Care Unit database.