Autonomous Detection of Cardiac Ailments Using ECG Signals and Random Forest Classifier

The ECG signal is a vital resource in diagnosing various cardiac ailments, providing crucial information for accurate decision-making re-‎garding different types of heart diseases. To achieve autonomous detection of cardiac ailments, several strategies have been proposed to ‎extract critical features from the ECG signal with utmost precision. In this study, a state-of-the-art methodology is introduced for the auto-‎matic detection of cardiac ailments. The proposed methodology encompasses three main steps: pre-processing, feature extraction, and classi-‎fication. In the pre-processing step, a Butterworth third-order band-pass filter is utilized to refine the ECG signal. For feature extraction, a ‎four-level maximal overlap discrete wavelet packet transform (MODWPT) technique is employed, utilizing the symlet wavelet as the mother ‎wavelet. In the final stage of classification, five supervised machine learning algor2 ns are applied to classify the considered three cardiac ‎ailments from the MIT-BIH database: Arrhythmia, Congestive Heart Failure, Atrial Fibrillation, and Normal Sinus Rhythm. The algorithms ‎used include Support Vector Machine (SVM), K-nearest Neighbour (KNN). Naive Bayes (NB), Decision Tree (DT), and Random Forest ‎‎(RF). These classifiers yield overall accuracies of 90.83%, 90.56%, 90.28%, 91.39%, and 91.94% respectively. Based on the results, it is ‎evident that the Random Forest classifier exhibits superior accuracy among the proposed methodology's classifiers for the multiclass classi-‎fication of cardiac ailments‎.