Title: Identification of important biomarkers for detection of chronic kidney disease using feature selection and classification algorithms

Authors: E. Sivasankar; R. Pradeep; S. Sivanandham

Addresses: Department of Computer Science and Engineering, National Institute of Technology, Tiruchirappalli, India ' Department of Computer Science and Engineering, National Institute of Technology, Tiruchirappalli, India ' Department of Computer Science and Engineering, National Institute of Technology, Tiruchirappalli, India

Abstract: Health informatics plays a critical role to discover, innovate and reform the healthcare system towards betterment by fostering collaboration between patients' data and medical practices. Development of easy, accurate and convenient methods for detection of diseases is a growing field of study in health informatics because of its easy scalability and affordability. In this paper, we have tried to identify important biomarkers for the detection of chronic kidney diseases. We have used filter- and wrapper-based feature selection techniques and four different classification methods to analyse and interpret data to help us diagnose CKD faster. We have reported the top attributes and their predictive accuracy in detecting CKD and support vector machine with Chi-squared feature selection method provided us with the best accuracy.

Keywords: chronic kidney disease; CKD; feature selection; classification; detection predictive model; biomarkers for CKD; risk factors; machine learning; Glomerular filtration rate; GFR; nephrology.

DOI: 10.1504/IJMEI.2019.104981

International Journal of Medical Engineering and Informatics, 2019 Vol.11 No.4, pp.368 - 385

Received: 12 Jul 2017
Accepted: 21 Mar 2018
Published online: 10 Feb 2020 *

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