CLASSIFICATION METHODS IN PROBLEMS OF DECISION-MAKING IN MEDICINE

Abstract

The results of comparison of classification methods in medical diagnostics problems are presented on the example of outcomes of coronary artery bypass graft surgery in acute myocardial infarction. We consider the probability of an outcome for an operation with the use of various techniques of myocardium protection: blood cardioplegia or cardioplegia using the Custodiol. The article compares simulation results obtained using such classification methods as logistic regression, decision trees and support vector machine. As compared criteria were used: contingency table, Area Under Curve, accuracy, completeness, F-measure, the Matthews correlation coefficient. As a result, it was revealed that some models are more preferable for use as a mathematical core when creating decision support information systems for choosing the method of cardioplegia.