Identifying and Comparing Influential Features in Diagnosing Heart Disease Using Machine Learning Methods

The prediction of cardiac disease helps experts make precise and more accurate decisions, relating to patients’ health. Therefore, the use of machine learning (ML) could be an answer to cut back and perceive the symptoms regarding heart disease. The aim of this article is the proposal of a dimensionality reduction technique and finding features of cardiopathy by applying some feature selection methods. The reason behind various strategies for feature selection is to assist specializations to settle on the most effective technique concerning to their want and therefore the performance of the method. The information which was used for this work was obtained from the UCI Machine Learning Repository known as “heart disease”. The dataset contains seventy-four features and a label that we have a tendency to valid by six ML classifiers. Soft-voting and random forest entropy had the best accuracy, with regarding ۹۱% and ۹۰%. From the analysis, features were derived of anatomical and physiological relevance, akin to cholesterol, max rate heart achieved, ST-depression and age. The application of these two methods directly from the information computed lower results and would need bigger dimensionality to improve the results.

کلیدواژه ها:

Heart disease ، Feature selection ، Machine learning ، Deep learning ، Early detection