Model-based clustering approach for clustering of heart disease patients based on risk factors

Abstract

Cardiovascular disease causes the most non-communicable disease deaths or 17.9 million people each year. One of the efforts that can be made to reduce the threat of heart disease death is the grouping of heart disease patients based on their risk factors. The risk factors are age, blood pressure, cholesterol, and maximum heart rate. The data in this study is continuous data with mixed distribution and cluster uncertainty. This research suggests a model-based clustering approach based on finite mixture models. This approach assumes that the data is generated by a mixture of probability distributions, where each distribution represents different clusters with different parameters. Model-based clustering allows each individual to have a probability value to enter another cluster, making it suitable when there is cluster uncertainty and determining the number of clusters can be done automatically. Therefore, clustering is performed using model-based clustering to automatically determine the optimal number of clusters and identify cluster uncertainty. Parameters of the model are estimated using the Expectation Maximization (EM) algorithm. The best model selection is determined based on the Bayesian Information Criterion (BIC) values. The clustering results obtained the best model EEI with the optimal number of clusters as many as two Clusters.