Recursive subspace based feature selection approach for early risk prediction of chronic disease in patients. (4th April 2022)

• Record Type: Journal Article
• Authors: Hegde, Sandeepkumar; Mundada, Monica R.
• Is Part Of: International journal of bioinformatics research and applications. Volume 18:Number 1/2(2022)
• Page Start: 4
• Page End: 15
• Publication Date: 2022-04-04
• Subjects: feature subspace -- covariance -- eigenvalue -- machine learning -- chronic disease -- diabetes mellitus -- logistic regression -- sigmoid activation -- grid hyperparameter -- projection matrix; Bioinformatics -- Periodicals; 570.285
• Languages: English
• ISSNs: 1744-5485

Abstract

Nowadays chronic disease is considered as a threat to human health. These diseases persist for a longer period of time and responsible for over 70% of death and disability all over the world. Healthcare findings and data are multidimensional in nature. Feature selection is considered as part of the pre-processing stage which is applied to the data with a higher dimension to lessen the features which impact the prediction of the disease. In this paper, a novel recursive subspace based feature selection (RSFS) algorithm is proposed. The feature subspace is obtained recursively by computing the covariance matrix and eigenvalue pairs. The experiment is conducted using a diabetes data set obtained from national institute of diabetes. Experimental results are compared with the chronic disease dataset of variant size with existing feature selection algorithms. The outcome demonstrates that the proposed RSFS algorithm performs better than existing techniques in terms of accuracy.