Machine Learning Based Classification Algorithms Performance Analysis for Heart Disease Prediction

Abstract

Recent study reveals that the mortality rate due to chronic diseases like heart disease is increasing year on year. Predicting heart disease at an early stage is posing a challenge to the healthcare industry due to multiple contributory factors like high blood pressure, uncontrolled cholesterol, obesity, sedentary lifestyle, smoking, alcohol consumption, etc. An accurate and effective diagnosis of heart disease at an early stage can prevent fatal complications such as heart attacks and strokes significantly. This research will not only help the medical fraternity, medico research scientists, and insurance agencies to assess the probability of heart disease but also help the common man to prevent hospitalization and reduce the expenses for the diagnosis significantly. In the past, multiple studies have been conducted on heart disease prediction using regular human vital parameters. We have expanded the research with family hereditary data of the person and by effectively using this feature we have evaluated model performance changes. We have used machine learning classification algorithms like Logistic Regression, KNN, Naive Bayes, and Decision Tree along with ensemble techniques like Random Forest with boosting algorithms like Ada Boost, XG Boost, etc. We evaluated the model performance with various metrics like precision, F1-score, and recall with more importance to the accuracy of the prediction.