Evaluation of Classification Techniques for Arrhythmia Screening of Astronauts. (2016)

• Record Type: Journal Article
• Authors: Deepthi, S.; Ravikumar, Aswathy; Nair, R. Vikraman
• Is Part Of: Procedia technology. Volume 24(2016)
• Page Start: 1232
• Page End: 1239
• Publication Date: 2016
• Subjects: Majority Voting -- Bagging -- Dagging -- DECORATE -- Support Vector Machine -- K-Nearest Neighbor -- Classification and Regression Tree -- Discriminant Analysis -- Naïve Bayes; Technology -- Congresses; Technology -- Periodicals; Engineering -- Congresses; Engineering -- Periodicals; Engineering; Technology; Conference proceedings; Periodicals; 605
• DOI: 10.1016/j.protcy.2016.05.099
• Languages: English
• ISSNs: 2212-0173

Abstract

Abstract: Arrhythmia is the major cause of cardiovascular events during space flight. Even though a number of physical tests are conducted to diagnose the disease, in most of the cases the issue remains undetected because of the hidden problems which cannot be pinpointed with regular physical tests. A computation system which can assist in analyzing hidden patterns of physical test is proposed which makes use of data mining and machine learning as the underlying approaches. The present study attempts to evaluate the performance of different individual classifiers such as Naïve Bayes, Support Vector machine (SVM), Classification and Regression Tree (CART), Linear Discriminant Analysis (LDA) and k-nearest neighbor (k-NN). Then the performance of these classifiers is compared with different ensemble techniques such as Majority Voting, Bagging, Dagging and DECORATE (Diverse Ensemble Creation by Oppositional Relabeling of Artificial Training Examples). The performance of the proposed classification methods is analyzed by considering different criteria such as accuracy, sensitivity and specificity. The result shows that among the individual classifiers implemented, k-nearest neighbor is having highest accuracy of around 84.44% only. But Majority Voting, which is an ensemble technique, is having the highest accuracy of 91.11% which is better than the individual classifier.