A Framework Design for Heart Failure Detection: Analyzes on Features and Hybrid Classifiers

The detection of heart failure is a vital and complicated issue that is needed to be analyzed comprehensively. On the basis of medicine, different tests and various scan techniques are utilized to efficiently make a decision. On the basis of machine learning, two phenomena come into prominence: 1-Qalitative data, 2-Framework design to detect the necessary information among the data.In this paper, an efficient framework is proposed to reveal the heart failure on the specific data. Three optimized classifiers were compared to assign the classification unit of framework. Manuel selection and filter based-feature ranking methods were considered to determine the necessary information and to reveal the heart failure. In experiments, two-fold cross validation was utilized as the test method to force the classifiers, and seven metrics based-comparisons were realized to objectively choose the features and classifiers. Consequently, the best framework achieved remarkable scores of 86.62% (accuracy), 83.01% (AUC), 72.92% (sensitivity), 93.10% (specificity), 82.39% (g-mean), 83.33% (precision) and 77.78% (f-measure) for survival prediction on heart failure clinical records.