Detecting Heart Abnormality using ECG with CART

Cardiovascular disease (CVD) is the leading cause of global deaths. Electrocardiogram (ECG or EKG) is the most widely used first line clinical tool for checking electrical activity in the heart. Hence using ECG recordings to automatically identify arrhythmias accurately and efficiently can be an important tool for cardiologists. We use the UC Irvine (UCI) Machine Learning Repository containing an arrhythmia data set to implement a multinomial classification for different types of heart abnormalities. We show that a decision tree learning algorithm (already widely used in many medical diagnostics) is well suited for this application. We achieved 80% classification accuracy with our decision tree, which is relatively high compared to other studies.