Table 3.
Evaluation metric table of samples for voice signals in classifying tube feeding.
| Accuracy (%) | Sensitivity (%) | Specificity (%) | NPV (%) | PPV (%) | F1 | AUC | |
|---|---|---|---|---|---|---|---|
| Voice only | |||||||
| LR | 68.2 (64.3–72.1) | 65.7 (58.2–73.1) | 70.7 (65.4–75.9) | 67.8 (63.1–72.6) | 69.3 (65.0–73.7) | 0.67 (0.62–0.72) | 0.69 (0.64–0.74) |
| DT | 69.0 (64.5–73.5) | 62.0 (56.5–67.5) | 76.0 (67.2–84.8) | 66.6 (63.2–70.0) | 73.3 (66.1–80.6) | 0.67 (0.62–0.71) | 0.70 (0.65–0.75) |
| RF | 73.7 (70.2–77.1) | 70.7 (66.1–75.3) | 76.7 (70.5–82.8) | 72.5 (69.2–75.7) | 75.7 (70.8–80.6) | 0.73 (0.69–0.76) | 0.78 (0.73–0.82) |
| SVM | 69.7 (65.9–73.5) | 71.0 (66.9–75.1) | 68.3 (62.8–73.9) | 70.2 (66.4–74.0) | 69.4 (65.2–73.5) | 0.70 (0.67–0.74) | 0.68 (0.63–0.73) |
| GMM | 66.2 (61.3–71.0) | 64.7 (51.3–78.1) | 67.7 (60.1–75.2) | 67.5 (61.2–73.7) | 66.3 (61.1–71.5) | 0.64 (0.55–0.74) | 0.64 (0.55–0.72) |
| XGBoost | 74.8 (71.0–78.7) | 72.7 (64.3–81.0) | 77.0 (68.9–85.1) | 74.8 (69.5–80.2) | 76.8 (71.2–82.4) | 0.74 (0.69–0.79) | 0.78 (0.73–0.82) |
| Voice + clinical | |||||||
| LR | 77.2 (74.4–80.0) | 76.7 (67.1–86.2) | 77.7 (70.1–85.2) | 78.3 (72.5–84.2) | 78.7 (73.5–83.8) | 0.77 (0.73–0.81) | 0.82 (0.79–0.85) |
| DT | 74.5 (70.6–78.4) | 80.0 (72.9–87.1) | 69.0 (64.8–73.2) | 78.3 (72.6–84.0) | 72.1 (68.8–75.3) | 0.76 (0.71–0.80) | 0.75 (0.71–0.80) |
| RF | 79.7 (75.9–83.4) | 85.0 (78.6–91.4) | 74.3 (67.0–81.7) | 83.9 (78.9–89.0) | 77.5 (72.3–82.7) | 0.81 (0.77–0.84) | 0.84 (0.80–0.88) |
| SVM | 77.0 (73.8–80.2) | 84.3 (78.7–90.0) | 69.7 (64.1–75.2) | 82.3 (77.4–87.1) | 73.8 (70.4–77.2) | 0.79 (0.75–0.82) | 0.81 (0.77–0.84) |
| GMM | 73.2 (68.9–77.5) | 76.3 (69.5–83.1) | 70.0 (63.6–76.4) | 75.4 (69.4–81.3) | 72.1 (67.6–76.6) | 0.74 (0.70–0.78) | 0.75 (0.71–0.79) |
| XGBoost | 82.5 (78.0–87.0) | 88.7 (82.6–94.7) | 76.3 (71.4–81.3) | 87.6 (81.6–93.5) | 79.0 (74.9–83.1) | 0.83 (0.79–0.88) | 0.85 (0.82–0.89) |
Values are presented in mean (95% confidence interval). Values with bold-text represent the highest values among the models.
NPV negative predictive value, PPV positive predictive value, AUC area under curve, LR logistic regression, DT decision tree, RF random forest, SVM support vector machine, GMM Gaussian mixture model, XGBoost extreme gradient boosting.