Using machine learning to determine the correlation between physiological and environmental parameters and the induction of acute mountain sickness

Table 2 Binary mild AMS classification results

Classifier type	Sensitivity	Specificity	Accuracy	AUC
Decision Trees
Fine Tree	0.998	0.978	0.996	0.9999
Medium Tree	0.993	0.952	0.988	0.99
Coarse Tree	0.975	0.862	0.963	0.90
Discriminant Analysis
Linear Discriminant	0.977	0.730	0.946	0.98
Quadratic Discriminant	0.997	0.707	0.952	0.99
Logistic Regression Classifiers
Logistic Regression	0.978	0.858	0.965	0.99
Naive Bayes Classifiers
Gaussian Naive Bayes	0.983	0.498	0.886	0.96
Kernel Naive Bayes	0.990	0.766	0.960	0.99
Support Vector Machines
Linear SVM	0.981	0.858	0.967	0.99
Quadratic SVM	0.995	0.939	0.989	0.9999
Cubic SVM	0.997	0.967	0.994	0.9999
Fine Gaussian SVM	0.995	0.975	0.992	0.9999
Medium Gaussian SVM	0.995	0.914	0.985	0.9999
Coarse Gaussian SVM	0.974	0.895	0.966	0.98
Nearest Neighbor Classifiers
Fine KNN	0.997	0.972	0.994	0.99
Medium KNN	0.996	0.957	0.991	0.9999
Coarse KNN	0.977	0.866	0.965	0.99
Cosine KNN	0.996	0.940	0.990	0.9999
Cubic KNN	0.995	0.949	0.990	0.9999
Weighted KNN	0.997	0.970	0.994	0.9999
Ensemble Classifiers
Boosted Trees	0.998	0.984	0.997	0.9999
Bagged Trees	0.999	0.994	0.998	0.9999
Subspace Discriminant	0.970	0.795	0.951	0.97
Subspace KNN	0.997	0.959	0.993	0.9999
RUSBoosted Tree	0.999	0.929	0.991	0.9999

AUC: area under the receiver operating characteristic curve
The Bagged Trees yielded the highest sensitivity, specificity, accuracy, and AUC; and was bolded for that reason

ISSN: 1471-2105