Prediction of Heart Disease by Optimized Distance and Density-Based Clustering

Abstract

Prediction of cardiovascular disease is considered as one of the most significant subjects in the section of clinical data analysis. In the previous studies, the prediction of heart disease is done by applying the data mining methods. However, a limited number of research works are existed towards predicting cardiovascular disease. It is complex for choosing the perfect combination of significant features, which can enhance the prediction model performance. From data and ECG signals, the robust and automated detection of heart disease plays a major task for accurate and early medical diagnosis. This paper presents the viability analysis and the development of heart disease prediction using data mining as well as signal processing techniques. The proposed methodology adopts the optimized clustering approach by means of unique improvement in distance and density-based clustering. Here, K-Means clustering is used as the distance-based clustering and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is used as the density-based clustering. Initially, the prediction is focused on the data gathered from the standard clinical repository, in which the new weighted feature extraction is adopted. Once the informative features are extracted, it is subjected to hybrid clustering approach, in which the output of both clustering is considered for finding the final output. Here, the hybrid clustering is formed by merging the optimized K-means clustering and optimized DBCAN. In K-means clustering (KMC), the centroid is optimized, and in DBSCAN, value of $\varepsilon$ is optimized by the new variant of meta-heuristic algorithm called Modified Updating based Chicken Swarm optimization (MU-CSO). Similarly, the ECG signals is gathered, which is decomposed using Discrete Wavelet Transform (DWT). Further, the dimension reduced features of the decomposed signals are extracted by Principal Component Analysis (PCA), to which the new weighted feature extraction is adopted. The main objective of optimized clustering is to maximize the prediction accuracy. Finally, the efficiency of the proposed heart disease prediction model is proved by comparing the proposed clustering over the conventional models.