Lasso regression and random forest analysis on variables contributing to myocardial infarction

• Published in: E3S web of conferences Vol. 553; p. 5034
• Main Author: Zhang, Xiucheng Eric
• Format: Journal Article
• Language: English
• Published: EDP Sciences 2024
• ISSN: 2267-1242; 2267-1242
• DOI: 10.1051/e3sconf/202455305034

Heart attack is a dangerous medical emergency that affects millions of people worldwide, and many factors may lead to its onset, such as high blood pressure and cholesterol. Since the cause of heart attack is very multivariate, it is imperative to understand which factors contribute more than others to achieve better clinical outcomes for heart attack patients. This paper collects data from a dataset that contains 1319 samples of eight variables that have been associated with causing increased likelihood of heart attack and aims to investigate which of the eight variables contribute to heart attack more than others. Lasso regression is applied to eliminate least relevant predicting variables in the dataset, and Random Forest model is applied to the remaining predictors to rank their relative importance to the output. It is found that the predictors troponin and KCM are the strongest contributing risk factors to the output heart attack. The finding may prove to be helpful for physicians’ heart attack diagnosis process in the future.