Title: Feature selection for detection of stroke risk using relief and classification method

Authors: Yonglai Zhang; Yaojian Zhou; Wenai Song

Addresses: Software School, North University of China, Taiyuan, 030051, China ' Software School, North University of China, Taiyuan, 030051, China ' Software School, North University of China, Taiyuan, 030051, China

Abstract: The morbidity of stroke presents an evident growing trend in the world. Stroke also features high disability rate and high recurrence rate. Therefore, the key to risk detection lies in preventing the stroke. This study mainly aims to find the way of selecting the most important influence factor in many features because of numerous risk factors of stroke. A new hybrid feature selection model is proposed based on a wrapper algorithm. The most important features are extracted from the data. Afterwards, a classification model aiming at the ischemic stroke is established with the support vector machine and GSO (glow-worm swarm optimisation) algorithm for the risk detection of diseases. The result of the classification shows that our method displayed good performance in the detection of ischemic stroke. The new method can provide the technical support for the stroke screening of mass population, and establish a referable application framework for the prevention of cardiovascular disease.

Keywords: classification; stroke risk; stroke; feature selection; relief; support vector machine; SVM.

DOI: 10.1504/IJMIC.2019.101967

International Journal of Modelling, Identification and Control, 2019 Vol.32 No.1, pp.46 - 53

Received: 11 Jul 2018
Accepted: 23 Nov 2018
Published online: 02 Sep 2019 *

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