Classification of Arrhythmia Using Conjunction of Machine Learning Algorithms and ECG Diagnostic Criteria

This paper proposes a classification technique using conjunction of Machine Learning Algorithms and ECG Diagnostic Criteria which improves the accuracy of detecting Arrhythmia using Electrocardiogram (ECG) data. ECG is the most widely used first line clinical instrument to record the electrical activities of the heart. The data-set from UC Irvine (UCI) Machine Learning Repository was used to implement a multi-class classification for different types of heart abnormalities. After implementing rigorous data preprocessing and feature selection techniques,different machine learning algorithms such as Neural Networks, Decision trees, Random Forest, Gradient Boosting and Support Vector Machines were used. Maximum experimental accuracy of 84.82% was obtained via the conjunction of SVM and Gradient Boosting. A further improvement in accuracy was obtained by validating the factors which were important for doctors to decide between normal and abnormal heart conditions.The performance of classification is evaluated using measures such as confusion matrix, kappa-score, confidence interval, Area under curve (AUC) and overall-accuracy. Key–Words: Machine Learning, Arrhythmia Classification, ECG, Neural Networks, SVM, Gradient Boosting