Receiver operating characteristic ( ROC ) and other curves measuring discriminability of classifiers ’ ensemble for asthma

Purpose: The aim was studying the discriminability by ROC curves and gain charts for simple fixed combining of constituent classifiers, for asthma severity diagnosis, and also for bagging and boosting. Material and methods: ROC shows a performance over a range of relative costs and probabilities a priori. Area under ROC curve (AUC) is the measure of separability of two probability distributions, for example of classifying functions. We examined ROC curves of different discriminant methods such as logistic regression, classification trees and neural networks. Next we combined these constituent classifiers and compared the obtained curve with curves of constituent classifiers. The analogous analysis was made on other methods of classifiers’ ensemble: bagging and boosting. Besides ROC-in the same way we examined also another curves, measuring discriminability cumulative and non-cumulative lift charts. Social and simple clinical data of 439 patients from three groups of children, hospitalized at the Institute of Pulmunology in Rabka, were used to find classification functions for existing and severity of asthma. We studied also two-group classification problems: asthmatic and non asthmatic children to elaborate automatic predicting of asthma. Results: We found out features with biggest discriminant properties in the differentiation groups of existing and severity of asthma. The improvement of performance after combining classifiers was proved by examining errors of classification and curves measuring discriminability. Conclusions: Performance of ensemble method can be visualized in one graph and compared with joint graph of constituent classifiers.