Multi-Target Classification for Predicting Myocardial Infarction Complications: A Comparative Analysis of Various Machine Learning Techniques

Abstract

Myocardial Infarction (MI) poses a significant challenge due to its potential for severe complications. Accurate identification of these complications is essential for improving prognosis. This paper investigates the application of a multi-label machine learning approach to classify complications arising from MI. These complications include heart failure, arrhythmias, and cardiogenic shock. The study uses patient datasets, features such as demographic information, clinical history, and biomarker levels. By employing advanced machine learning models, including binary relevance, classifier chains, label powerset, and RAKEL-D, the research aims to enhance the predictive accuracy of MI complication diagnosis. The findings demonstrate that the RAKEL-D model outperforms other classifiers in accurately identifying concurrent complications. It shows an f1 score between 0.615 to 0.64 across five-fold, around 0.054 to 0.057 hamming loss, 0.62 Jaccard, and 0.442 accuracy.