ECG-Based Arrhythmia Classification in Students Using Random Forest: A Case Study with Class Imbalance Analysis

Abstract

Arrhythmia is a heart rhythm disorder that can indicate a student’s heart health status. This research aims to develop a Random Forest model to classify arrhythmia in students based on ECG signals. ECG data was collected from 100 students at SMK Swasta Teladan Sumatera Utara 2 after learning activities. The extracted signal features include RR interval, PR interval, QRS duration, QT interval, ST segment, beats per minute (BPM) and R/S ratio. Data labeling was carried out manually by the researchers based on the range of ECG feature values that had been determined by the doctor for each class: Normal, Abnormal, Potential Arrhythmia and Very Potential Arrhythmia. The dataset is divided into 70% for training and 30% for testing. SMOTE is applied to address class imbalance. The model achieved 80% accuracy with the best performance in normal class with precision, recall and f1-score  of  94%. However, no samples were identified for Potential Arrhythmia class, as there were no extracted feature values that met the criteria set by the doctor, so model could neither learn nor make predictions for this category, even after applying balancing methods such as SMOTE. For further research, based on these findings, it highlights the need for balanced class representation and expert-guided labeling to improve the performance of ECG -based arrhythmia classification.