Hypertension Type Classification Using Hierarchical Ensemble of One-Class Classifiers for Imbalanced Data

Springer papers

Authors

Bartosz Krawczyk and Michal Wozniak

Abstract

The paper presents the research on the computer support system which is able to recognize the type of hypertension. This diagnostic problem is highly imbalanced, because only ca. 5% of patient suffering from hypertension are diagnosed as secondary hypertension. Additionally the secondary hypertension could be caused by several disorders (in our work we recognize the five most popular reasons) which require strikingly different therapies. Thus, appropriate classification methods, which take into consideration the nature of the decision task should be applied to this problem. We decided to employ the original classification methods developed by our team which have their origin in one-class classification and the ensemble learning. They quality was confirmed in our previous works. The accuracy of the chosen classifiers was evaluated on the basis of the computer experiments which were carried out on the real data set obtained from the hypertension clinic. The results of the experimental investigations confirmed usefulness of the proposed, hierarchical one-class classifier ensemble and could be applied in the real medical decision support systems.

Keywords

classifier ensemble pattern classification one-class classifier imbalanced data hypertension