Table 2 Comparison of Machine Learning Algorithms for Prediction of Delirium in Two Overlapping Feature Sets (Ranking based on AUC)
Algorithms | Accuracy | Precision | Recall | F1 | AUC (95% CI) | AP (95% CI) | ECE |
|---|---|---|---|---|---|---|---|
Full feature set (q = 31) |  | ||||||
Random Forest Classifier | 0.83 | 0.80 | 0.86 | 0.81 | 0.92 (0.91–0.92) | 0.80 (0.78–0.82) | 0.10 |
Gradient Boosting Decision Tree | 0.84 | 0.80 | 0.80 | 0.80 | 0.90 (0.89–0.91) | 0.79 (0.76–0.81) | 0.05 |
Support Vector Machine Classifier | 0.83 | 0.79 | 0.81 | 0.80 | 0.83 (0.81–0.84) | 0.68 (0.66–0.71) | 0.15 |
Logistic Regression | 0.76 | 0.71 | 0.74 | 0.72 | 0.80 (0.78–0.82) | 0.64 (0.61–0.67) | 0.06 |
Gaussian Naive Bayes | 0.76 | 0.72 | 0.76 | 0.73 | 0.79 (0.77–0.80) | 0.57 (0.55–0.59) | 0.08 |
K-nearest Neighbors Classifier | 0.76 | 0.74 | 0.60 | 0.61 | 0.78 (0.76–0.80) | 0.55 (0.52–0.58) | 0.14 |
Perceptron | 0.78 | 0.73 | 0.73 | 0.73 | 0.77 (0.75–0.78) | 0.60 (0.57–0.62) | 0.23 |
Selected feature set (q = 19) |  | ||||||
Random Forest Classifier | 0.78 | 0.73 | 0.73 | 0.73 | 0.86 (0.85–0.88) | 0.73 (0.70–0.75) | 0.14 |
Gradient Boosting Decision Tree | 0.76 | 0.74 | 0.60 | 0.61 | 0.83 (0.82–0.84) | 0.61 (0.59–0.64) | 0.13 |
Logistic Regression | 0.83 | 0.80 | 0.86 | 0.81 | 0.76 (0.74–0.77) | 0.61 (0.58–0.63) | 0.13 |
K-nearest Neighbors Classifier | 0.83 | 0.79 | 0.81 | 0.80 | 0.72 (0.70–0.74) | 0.50 (0.47–0.53) | 0.11 |
Gaussian Naive Bayes | 0.76 | 0.71 | 0.74 | 0.72 | 0.70 (0.68–0.72) | 0.50 (0.48–0.53) | 0.12 |
Perceptron | 0.76 | 0.72 | 0.76 | 0.73 | 0.69 (0.67–0.71) | 0.50 (0.47–0.54) | 0.31 |
Support Vector Machine Classifier | 0.84 | 0.80 | 0.80 | 0.80 | 0.64 (0.62–0.67) | 0.52 (0.49–0.55) | 0.14 |