The area under the ROC is a convenient way of comparing classifiers.
In general, values for overall power demonstrated superior performance for the DLDA classifier compared to the ASD classifiers in all scenarios.
The results show that in average, KNN is the least performing classifier compared to Bayes classifier with Gaussian and MD model as PDF estimators.
We compared all classifiers according to a number of the correctly classified chemicals and the classifiers accuracies.
Two fuzzy classifiers, a Bayesian classifier and a linear classifier, are designed for vocal folds damage detection based on human vowel voices /a:/ and /i:/ only, and the fuzzy classifiers are compared against the Bayesian classifier and linear classifier.
Regarding compared learning classifiers, NMSC outperforms the others with respect to testing performance, stabilization, and consistency.
Compared with classifiers with higher ranks in F1 (Table 1), the results appear to suggest that classifiers with high PPV achieve better hit ratios.
In order to compare the classifiers by using just one error number instead of 17, we compute a weighted average for each classifier's time response.
We subsequently compare the recognition performances between the RVM classifiers and three other popular classification techniques, i.e., k-NN, Bayesian decision, and SVM.
The area under the curve gives a comprehensive figure-of-merit for prediction accuracy and can be shown to be a consistent but more sensitive measure than error rate for comparing classifiers, identifying performance differences between classifiers in cases where, evaluated solely by error rate, two classifiers would be deemed equivalent (Swets, 1988; Wang et al, 2006).
