Anomaly Detection in Heart Disease Using a Density-Based Unsupervised Approach

Abstract

Cardiovascular disease is one of the most common diseases in the modern world, which, if diagnosed early, can greatly reduce the damage to the patient. Diagnosis of heart disease requires great care, and in some cases, the process can be disrupted by human error. Machine learning methods, especially data mining, have gained international acceptance in almost all aspects of life, especially the prediction of heart disease. On the other hand, datasets related to heart patients have many biological features that most of these features do not have a direct impact on diagnosis. By removing redundant features from the dataset, in addition to reducing computational complexity, the accuracy of heart patients’ predictions can also be increased. This paper presents a density-based unsupervised approach to the diagnosis of abnormalities in heart patients. In this method, the basic features in the dataset are first selected based on the filter-based feature selection approach. Then, the DBSCAN clustering method with adaptive parameters has used to increase the clustering accuracy of healthy instances and to determine abnormal instances as cardiac patients. Partition clustering methods suffer from the selection of the number of clusters and the initial central points and are very sensitive to noise. The DBSCAN method solves these problems by creating density-based clusters, but the selection of the neighborhood radius threshold and the number of connected points in the neighborhood remains unresolved. In the proposed method, these two parameters are selected adaptively to achieve the highest accuracy for the diagnosis and prediction of heart patients. The results of the experiments show that the accuracy of the proposed method for predicting heart patients is approximately 95%, which has improved in comparison with previous methods.