In cardiology, determining whether an electrocardiogram (ECG) is normal or not is sometimes referred to as ECG classification. It is crucial for making a diagnosis and even for deciding whether or not surgery is necessary. Automating ECG classification and improving its accuracy is an active area of research. In this paper we derive fuzzy rules for ECG classification from ID3-induced decision trees. The rules are designed based on 106 ECG's and tested using a validation set of 48 ECG's selected by cardiologists. An ID3-generated decision tree designed using the same 106 ECG's and tested on the same 48 validation ECG's yields 73% correct classifications. On the other hand, the derived fuzzy rules, combined with an optimized defuzzification using the cascade correlation neural network, produce 100% correct classifications.
展开▼
