A Comparative Analysis of Machine Learning Algorithms for Heart Diseases Prediction

Abstract

The term heart disease refers to the range of conditions that impacts heart and its blood vessels in a negative manner. This has led to increase mortality rate worldwide and a major health issue in recent years and continues to have increasing statistics in future. Modern day lifestyles and choices have made it increasingly common, not only in aged population but also in younger adults. New generation advanced medicines are fortunately a breakthrough in assisting and curing the disease, but conventional techniques are only optimal to provide on the spot single data reading and result, it fails to predict the future circumstances and possible emergency which may be caused due to underlying issue, or provide with early diagnosis to facilitate prompt medical assistance. This is where Machine learning (ML) comes into the picture, delivering the required and necessary early analysis of cardiovascular disease (CVD) through validating various features among thousands of data in a dataset. This study has developed predictive models and used Cleveland dataset with 11 different features namely age, sex, cholesterol, resting heart rate, exercise-induced angina and other health-related indicators that make up this dataset.  There are four different predictive models that we will be using here to predict the health analysis of heart related disease, which are Logistic Regression, Random Forest, Support Vector Machines (SVM), K-Nearest Neighbor (KNN). All of the mentioned models are developed and used under Google colab platform. There are 4 different datasets consisting of Cleveland, Hungary, Switzerland and VA Long Beach which are preprocessed and cleaned and then being used and tested here with machine learning models to test its accuracy and other result aspects. The best result and performance among these four models tested across each of the datasets and having minimum spread (max-min) is of Support Vector Machine (SVM). with an F1 score of 0.7732.  This study will focus on developing multi predictive models to help pave a way to ensure early diagnosis and provide early treatment. Keywords— heart; disease; predictive; machine learning; models