Handling imbalanced class problem for the classification of hypertension in the coronary artery disease patients by using medical knowledge discovery process

Aim: The primary aim of this study is to estimate (classify) hypertension, one of the causes of mortality and morbidity increase in coronary artery disease patients, by applying Medical Knowledge Discovery Process with various risk factors. Because of the class imbalance problem of hypertension, which is dependent variable of the dataset used in the study, the development of a web-based software which uses various approaches to resolve this problem before the classification process and whose interface is Turkish is the second main aim of this study. Material and Method: The dataset used in the study consisted of records of 929 coronary artery patients with 149 (16%) hypertension and 780 (84%) non-hypertension. Classification of hypertension in coronary artery patients was done based on 8 independent variables. Various over-under sampling and both over and under sampling methods was used to handle the imbalanced class problem. As the classification methods, Multilayer Perceptron, Extreme Learning Machine and Support Vector Machine models were performed. Results: The best classification performance was obtained by the Support Vector Machine model after applying the DBSMOTE class balancing method. The accuracy, sensitivity, specificity, precision, f-measure and g-mean metrics of the relevant model were calculated as 0.99, 0.99, 0.99, 0.95, 0.97 and 0.97, respectively. Conclusion: Compared to the undersampling methods, the oversampling methods used in the study showed a positive contribution to the classification performance of the models. Hybrid Methods, Cost-Sensitive Learning Based Methods, Ensemble Learning Based Methods, Feature Selection Based Methods, which aren't included in the scope of this study but will be discussed in further studies, can be suggested to readers for more robust and consistent results. Key Words: Imbalanced class problem, over and under sampling methods, hypertension, medical knowledge discovery process, coronary artery disease.