Multiclass Classification of Cardiac Arrhythmia Using Improved Feature Selection and SVM Invariants

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Arrhythmia poses significant health risks if untreated.
• Early detection of arrhythmias is vital for patient survival.
• Classifying various arrhythmia subtypes is essential for effective treatment.

Purpose of the Study:

• To classify patients into 16 subclasses, including absence of disease and 15 arrhythmia subtypes.
• To evaluate the effectiveness of Support Vector Machine (SVM) based approaches for arrhythmia detection.
• To compare SVM performance against other machine learning classifiers.

Main Methods:

• Utilized a dataset from the UCI Machine Learning Data Repository.
• Reduced high-dimensional features using a wrapper-based feature selection technique.
• Employed multiclass classification with SVM variants: One-Against-One (OAO), One-Against-All (OAA), and Error-Correction Code (ECC).

Main Results:

• The One-Against-One (OAO) SVM method demonstrated superior performance.
• Achieved an accuracy of 81.11% with an 80/20 data split.
• Reached an accuracy of 92.07% with a 90/10 data split.

Conclusions:

• The OAO-SVM approach is highly effective for classifying multiple arrhythmia subtypes.
• This method offers a promising tool for early and accurate arrhythmia diagnosis.
• SVM-based classification, particularly OAO, shows significant potential in cardiovascular disease detection.