Table 4.

Comparison of different measures of prediction accuracy and number of features used across different classifiers trained using all of the survey questions. Reported results are over a 5 × 5 nested cross-validation with mean ± standard error over the outer folds. Bold entries denote highest accuracy, sensitivity, specificity, and lowest number of features used.

Classifier	Accuracy	Sensitivity	Specificity	Features Used
Logistic Regression	0.653 ± 0.034	0.407 ± 0.061	0.800 ± 0.067	37.4 ± 10.4
Random Forest	0.659 ± 0.025	0.513 ± 0.086	0.744 ± 0.038	70.6 ± 15.0
Gradient Boosting	0.666 ± 0.027	0.402 ± 0.092	0.822 ± 0.021	38.0 ± 18.3