Enhancing ECG Arrhythmia Classification via Stacked Ensemble Learning: A Machine Learning Approach

A person’s health can be viewed from the condition of their heart health. Arrhythmia is a common heart disorder in society. Arrhythmias occur as a complication of the heart’s normal rhythm functions. The most recent global epidemiological data reported by the Ministry of Health of the Republic Indonesia suggests that there is a 1.5%-5% prevalence rate for arrhythmias in a given population. An Electrocardiogram (ECG) provides one means of diagnosing aberrations and detecting an arrhythmia disorder. Electrocardiogram signals are classified based on the pulse according to the AAMI standard. The use of Ensemble Machine Learning helps medical personnel classify arrhythmias through ECG signals more precisely and accurately. Stacking is one of the Ensemble Learning frameworks that uses a combination of predictions from several base learners into a meta-learner to obtain a better final level of accuracy. The Single-Layer Stacking, Multi-Layer Stacking and Blending algorithms will classify arrhythmias by identifying patterns based on their pulses. This study tested several models, namely Single-Layer Stacking, Multi-layer Stacking, and Blending to determine the optimal model based on accuracy metrics. The results indicated that Multi-layer Stacking achieved the highest accuracy of 85.20%. When viewed from other ensemble techniques, Stacking offers a more flexible solution for handling complex data and provides accurate prediction results. The conclusion of this study is that Multi-layer Stacking is the most effective technique for arrhythmia classification.