Binary Multi-Verse Optimization (BMVO) Approaches for Feature Selection

Rahul Hans

doi:10.9781/ijimai.2019.07.004

Binary Multi-Verse Optimization (BMVO) Approaches for Feature Selection

2020-01-01

18

OA

AI

DOI：10.9781/ijimai.2019.07.004

Original

Original request for help

Save

Abstract

En 中文

Multi-Verse Optimization (MVO) is one of the newest meta-heuristic optimization algorithms which imitates the theory of Multi-Verse in Physics and resembles the interaction among the various universes. In problem domains like feature selection, the solutions are often constrained to the binary values viz. 0 and 1. With regard to this, in this paper, binary versions of MVO algorithm have been proposed with two prime aims: firstly, to remove redundant and irrelevant features from the dataset and secondly, to achieve better classification accuracy. The proposed binary versions use the concept of transformation functions for the mapping of a continuous version of the MVO algorithm to its binary versions. For carrying out the experiments, 21 diverse datasets have been used to compare the Binary MVO (BMVO) with some binary versions of existing metaheuristic algorithms. It has been observed that the proposed BMVO approaches have outperformed in terms of a number of features selected and the accuracy of the classification process.

Keywords:

Machine Learning

Feature Selection

K-Nearest Neighbors

Binary Multi-Verse Optimization

Binary Multi-Verse Optimization (BMVO) Approaches for Feature Selection

Abstract

Journal

Organization

Cited Papers