Özet:

Aim: Coronary artery disease is one of the most fatal diseases in the the society. Early diagnosis and treatment of coronary artery disease plays an important role in reducing the number of deaths. In this study, it is aimed to classify coronary artery disease by Stacking based ensemble learning methods. Material and Methods: The study was obtained from the data of 244 patients with coronary artery disease and 116 individuals without coronary artery disease who were treated in Kahramanmaras Sutcu Imam University Health Practice and Research Hospital. The data were obtained retrospectively. The data set consists of 15 predictor variables and 1 dependent variable. In the classification process, Naive Bayes, Sequential Minimal Optimization, Random Forest classifiers and Stacking ensemble learning method were applied. A 10-fold cross validation method was applied to the model. Accuracy, sensitivity, specificity, F-measure and AUC metrics were applied to evaluate the performance of classifiers. The most essential variables in predicting coronary artery disease have been determined. Results: ACC = 0.774, SEN = 0.888, SPE = 0.719, F = 0.718 and AUC = 0.913 values were obtained with the Naive Bayes classifier in the study. ACC = 0.883, SEN = 0.733, SPE = 0.955, F = 0.802 and AUC = 0.844 were obtained with the SMO classifier. ACC = 0.908, SEN = 0.853, SPE = 0.934, F = 0.857 and AUC = 0.957 were obtained with Random Forest classifier. ACC = 0.933, SEN = 0.905, SPE = 0.946, F = 0.897 and AUC = 0.959 values were obtained with the stacking ensemble learning method. BUN, MPV, Age, AST and Monocyte variables were determined as the most essential factors in the classification of coronary artery disease, respectively. Coclusion: Stacking ensemble learning method provided the highest accuracy performance in the classification of coronary artery disease. Stacking ensemble learning method gives successful results in the classification of coronary artery diseases.

Anahtar Kelimeler: