Comparative Analysis of Supervised Machine Learning Algorithms for Predicting Cardiovascular Disease

Abstract

Heart disease is one of the most dangerous and largest killers in the world. Identifying heart disease early has a vast positive effect on patient outcomes and their quality of life. In this research, we try to identify heart disease using machine learning (ML) algorithms. ML algorithms have the highest probability of success if they work on a data set with extensive and diverse information about the given problem - in this case heart disease. There are multiple types of ML algorithms to test, so we can try many different ones on the data, making the results more precise. Even if there are many different algorithms to test, each machine-learning solution can yield different results depending on the dataset used and our target goals. The main goal of this study is to find the differences between individual ML algorithms being used in our specific case: which ML algorithm, or combination of algorithms, is appropriate to detect heart disease with high accuracy? The ML algorithms used in this research are the Naive Bayes Classifier, the Random Forest classifier, and the Support Vector Machine (SVM) algorithm. Furthermore, this study generates insights into these ML algorithms – if a particular algorithm’s model performs better than another on the dataset, analyzing this difference can help us understand what makes the model more suitable for diagnostic screening. Changing models’ hyperparameters or their pre-processing techniques allows for a more robust and reliable model that can be readily incorporated into a healthcare environment.