Evaluation of ECG: comparison of decision tree and fuzzy rules induction

Abstract

This paper compares two different approaches to computer-aided analysis of ECG signals. ECG records are preprocessed by the wavelet transform, and the machine learning method of decision trees and fuzzy rules induction are used for classification. The wavelet transform allows good localisation of QRS complexes, P and T waves in time and amplitude. The average accuracy of detection of all events is above 87 per cent. For learning and further classification we use Quinlan's See5 application and FURL (FUzzy Rule Learner). We used the MIT-BIH database for experiments. Diverse settings of the parameters for decision tree generation (tree pruning, attribute selection, class sets) were examined. Two datasets and diverse settings of fuzzysets were examined as well.