Binary dragonfly optimization for feature selection using time-varying transfer functions

Mafarja, M, Aljarah, I, Heidari, A, Faris, H, Fournier-Viger, P, Li, X and Mirjalili, S 2018, 'Binary dragonfly optimization for feature selection using time-varying transfer functions', Knowledge-Based Systems, vol. 161, pp. 185-204.


Document type: Journal Article
Collection: Journal Articles

Title Binary dragonfly optimization for feature selection using time-varying transfer functions
Author(s) Mafarja, M
Aljarah, I
Heidari, A
Faris, H
Fournier-Viger, P
Li, X
Mirjalili, S
Year 2018
Journal name Knowledge-Based Systems
Volume number 161
Start page 185
End page 204
Total pages 20
Publisher Elsevier BV
Abstract The Dragonfly Algorithm (DA) is a recently proposed heuristic search algorithm that was shown to have excellent performance for numerous optimization problems. In this paper, a wrapper-feature selection algorithm is proposed based on the Binary Dragonfly Algorithm (BDA). The key component of the BDA is the transfer function that maps a continuous search space to a discrete search space. In this study, eight transfer functions, categorized into two families (S-shaped and V-shaped functions) are integrated into the BDA and evaluated using eighteen benchmark datasets obtained from the UCI data repository. The main contribution of this paper is the proposal of time-varying S-shaped and V-shaped transfer functions to leverage the impact of the step vector on balancing exploration and exploitation. During the early stages of the optimization process, the probability of changing the position of an element is high, which facilitates the exploration of new solutions starting from the initial population. On the other hand, the probability of changing the position of an element becomes lower towards the end of the optimization process. This behavior is obtained by considering the current iteration number as a parameter of transfer functions. The performance of the proposed approaches is compared with that of other state-of-art approaches including the DA, binary grey wolf optimizer (bGWO), binary gravitational search algorithm (BGSA), binary bat algorithm (BBA), particle swarm optimization (PSO), and genetic algorithm in terms of classification accuracy, sensitivity, specificity, area under the curve, and number of selected attributes. Results show that the time-varying S-shaped BDA approach outperforms compared approaches.
Subject Optimisation
Neural, Evolutionary and Fuzzy Computation
Keyword(s) Binary dragonfly algorithm
Classification
Feature selection
Optimization
Transfer functions
DOI - identifier 10.1016/j.knosys.2018.08.003
Copyright notice © 2018 Elsevier B.V.
ISSN 0950-7051
Versions
Version Filter Type
Citation counts: TR Web of Science Citation Count  Cited 25 times in Thomson Reuters Web of Science Article | Citations
Scopus Citation Count Cited 0 times in Scopus Article
Altmetric details:
Access Statistics: 18 Abstract Views  -  Detailed Statistics
Created: Thu, 21 Feb 2019, 12:10:00 EST by Catalyst Administrator
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us