Detecting Arrhythmias Using ECG Signals with Machine Learning Techniques

This paper presents a comparative study on the effectiveness of various feature selection methods and machine learning algorithms for arrhythmia prediction using electrocardiogram (ECG) signals. ECGs are crucial for diagnosing abnormal heart rhythms, known as arrhythmias, which arise from changes in the heart's electrical impulse patterns. The study focuses on two primary feature selection techniques: the Sequential Feature Selection (a wrapper method) and the Pearson Correlation (a filter method). These methods were applied to a real-world medical dataset to assess the impact on machine learning model performance. Our approach involved data training and testing through methods such as Cross-Validation Partition and an 80-20 training-testing split, iterating through the methods to find optimal conditions. A number of machine learning models were tested, including Naive Bayes Classifiers, K-Nearest Neighbors (KNN), Support Vector Machines (SVM), and Ensemble Methods. Initial tests used a generic Naive Bayes implementation, which was later evaluated against reduced feature sets generated by Sequential Feature Selection and Pearson Correlation. The performance of these models was measured using metrics such as accuracy, balanced accuracy, sensitivity, and specificity. Results showed that the Naive Bayes Classifier achieved the highest accuracy, and both Naive Bayes and SVM exhibited high specificity at 94%. This study underscores the value of targeted feature selection in enhancing the predictive accuracy and efficiency of machine learning models in medical diagnostics, particularly in predicting arrhythmias from ECG data.