A random walk Grey wolf optimizer based on dispersion factor for feature selection on chronic disease prediction. (15th November 2022)

• Record Type: Journal Article
• Authors: Preeti, ; Deep, Kusum
• Is Part Of: Expert systems with applications. Volume 206(2022)
• Publication Date: 2022-11-15
• Subjects: Feature selection -- Medical data -- Grey Wolf optimizer -- Classification problem; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2022.117864
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: Improvised Grey wolf optimizer for feature selection on medical data. Proposed method ranks best over other Nature Inspired Algorithms. Identification of significant features of chronic disease data. Prediction of the output class for each data using significant features. Abstract: In the field of Chronic disease prediction, identifying the relevant features plays an important role for early disease diagnosis. With a high dimensionality of data, search for an adequate subset feature increases exponentially and is categorised as NP hard problem. Nature Inspired Algorithm (NIA) have been famous to tackle this problem by finding an optimum solution in a reasonable amount of time. Grey wolf optimizer (GWO) is an emerging and powerful NIA used in wrapper feature selection method. It is well known for its flexibility, simplicity and efficient results. However, GWO have unsatisfactory results on local searching ability, and a slow convergence rate. To improve the local search and find a balance between exploration and exploitation, this paper proposes a Random Walk Grey Wolf Optimizer based on dispersion factor (RWGWO) approach to the feature selection problem. To demonstrate the methodology, a set of classification measures is evaluated and examined on eighteen different chronic disease data. For a fair comparison, RWGWO is compared with several recent state of the art method. Finding shows that RWGWO method ranks best over other NIAs and is able to drastically reduces the features size on each chronic disease data. Further, identification of significant set of features from each data is determined using obtained features of RWGWO. The significant features are able to enhance the classification accuracy of the presented data and solves the dimensionality reduction problem.