Hybrid Ensemble Feature Selection for Heart Disease Prediction System Using an NMF Hierarchical Clustering

Heart disease is the leading cause of death in the world over the past 10 years. Researchers have been using several data mining techniques to help health care professionals in the diagnosis of heart disease patients. It is essential to find the best fit clustering or classification algorithm that has greater accuracy on prediction in the case of heart disease. Heart disease patient database is datasets collected from Cleveland Heart Disease Dataset (CHDD) available on the University of California, Irvine (UCI) Repository. Since the database samples consist of huge attribute and selection of best attribute from this CHDD becomes very important for prediction accuracy. This paper presents a novel Hybrid Feature Selection Ensemble (HFSE) framework generalising the ensemble approaches so that it can be combined with many feature selection (FS) approaches based on Hybrid Genetic and Particle Swarm Optimization (HGPSO), Hybrid Particle Swarm Optimization with Hidden Markov Model (HPSO-HMM) algorithm for computer aided disease diagnosis scheme. The aim is to design swarm intelligence methods for dimensionality reduction of heart disease diagnosis. The designed HFSE framework is applied to CHDD. Originally there were 13 attributes involved in predicting the heart disease. In this work, HFSE is used to determine the attributes that contribute more towards the diagnosis. Thirteen attributes are reduced to seven attributes using HGPSO and HPSO-HMM. In this paper, an efficient approach, Non-negative Matrix Factorization with Hierarchical Clustering methods (NMF–HC) with Features Selected Data (FSD), is proposed for the intelligent heart disease prediction. The system designed in Matlab software can be viewed as an alternative for existing methods to distinguish of heart disease presence.