Skip to main content

Table 3 Performance of the different classifiers

From: Predicting asthma control deterioration in children

Performance of the decision stump classifier
Evaluation method Sensitivity Accuracy Specificity AUC PPVa NPVa
10-fold cross validation 67.2 % 73.4 % 74.9 % 0.710 38.1 % 91.0 %
testing on each patient’s last assessment 51.1 % 73.9 % 82.0 % 0.665 50.0 % 82.6 %
Performance of the six advanced classifiers measured by the 10-fold cross validation method
Classifier Sensitivity Accuracy Specificity AUC PPVa NPVa
Multiboost with decision stumps 73.8 % 71.8 % 71.4 % 0.761 37.1 % 92.4 %
Support vector machine 71.5 % 72.0 % 72.0 % 0.718 37.0 % 91.8 %
Deep learning 71.6 % 72.3 % 72.5 % 0.744 37.2 % 91.8 %
Naive Bayes 59.8 % 78.1 % 82.3 % 0.777 43.7 % 90.0 %
k-nearest neighbor 56.9 % 73.3 % 77.0 % 0.704 36.0 % 88.7 %
Random forest 48.3 % 75.8 % 82.0 % 0.662 37.9 % 87.5 %
Performance of the six advanced classifiers measured by the method of testing on each patient’s last assessment
Classifier Sensitivity Accuracy Specificity AUC PPVa NPVa
Multiboost with decision stumps 74.5 % 74.4 % 74.4 % 0.757 50.7 % 89.2 %
Support vector machine 70.2 % 73.3 % 74.4 % 0.723 49.3 % 87.6 %
Deep learning 68.1 % 72.2 % 73.7 % 0.738 47.8 % 86.7 %
Naive Bayes 44.7 % 73.9 % 84.2 % 0.783 50.0 % 81.2 %
k-nearest neighbor 48.9 % 73.9 % 82.7 % 0.773 50.0 % 82.1 %
Random forest 38.3 % 75.6 % 88.7 % 0.678 54.5 % 80.3 %
  1. aPPV positive predictive value; NPV negative predictive value