Title: Prediction of heart diseases using hybrid feature selection and modified Laplacian pyramid non-linear diffusion with soft computing methods

Authors: R. Suganya; S. Rajaram; A. Sheik Abdullah; V. Rajendran

Addresses: Department of Information Technology, Thiagarajar College of Engineering, Madurai, India ' Department of ECE, Thiagarajar College of Engineering, Madurai, India ' Department of Information Technology, Thiagarajar College of Engineering, Madurai, India ' Department of Cardiology, Covai Heart Foundation, Coimbatore, India

Abstract: The main objective of this paper is to predict the possibility of heart diseases at its early stages with less number of attributes. Pre-processing helps to improve sensor images by removing noise present in it. In the proposed work, a coupled modified diffusivity function is applied in Laplacian pyramid domain of an image, to eliminate noise and retains subtle features simultaneously. Our approach integrates anthropometric data and physiological data of heart diseases by proposing hybrid feature selection method for prediction of heart diseases using soft computing techniques. We ran experiments in neural network and SVM and proved that the neural network predicts 92% of accuracy and SVM predicts 97% of accuracy. The results show the proposed approach leads to a superior feature selection process in terms of sinking the number of variable required and an increased in classification accuracy for better prediction.

Keywords: hybrid feature selection; modified Laplacian pyramid; non-linear diffusion; neural network; support vector machine; heart diseases; sensor data.

DOI: 10.1504/IJBET.2017.086550

International Journal of Biomedical Engineering and Technology, 2017 Vol.25 No.1, pp.30 - 43

Received: 25 May 2016
Accepted: 07 Jul 2016

Published online: 29 Aug 2017 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article