Enhanced Global Best Particle Swarm Classification

Enhanced Global Best Particle Swarm Classification

Nabila Nouaouria (Department of Computer Science, UQAM, Montreal, Canada), Mounir Boukadoum (Department of Computer Science, UQAM, Montreal, Canada) and Robert Proulx (UQAM, Montreal, Canada)
DOI: 10.4018/ijssci.2014070101
OnDemand PDF Download:
List Price: $37.50
10% Discount:-$3.75


Particle Swarm Classification (PSC) is a derivative of Particle Swarm Optimization (PSO) based on the retrieval of the best particle positions corresponding to the centroids of classes. This paper addresses how the position update mechanisms impacts the accuracy of a global best PSC approach. The authors present two variants of the PSC algorithm with different position update mechanisms. In particular, the authors show how the combination of a good parameters tuning, a particle confinement to the search space and a biologically inspired wind dispersion mechanism for them improves the covering quality of search space and thus the classification accuracy of the basic global PSC algorithm. An experimental set up was realized and tested on five benchmark databases, leading to better recognition accuracies than those obtained with the previous PSC algorithm.
Article Preview

2. Pso-Based Classification Background

The roots of PSO lay in ethological metaphors of computing models (Poli, Kenndy & Blackwell, 2001; Kennedy, Eberhar, 2001; Abraham, Guo, Liu, 2006). Formally, a PSO algorithm is based on a swarm of M individuals or particles, each evolving in N-dimensional space with its coordinates representing a potential solution to a problem with N attributes. Its genotype consists of 2N parameters, the first half representing the coordinates of the particle in the search space and the second half the corresponding velocity components. A particle moves with adaptable velocity within the search space, and it constantly updates and retains the best position it ever reaches in memory. The best position reached by the swarm, and sometimes also in a neighborhood of particles, are also constantly updated and memorized.

Complete Article List

Search this Journal:
Volume 15: 1 Issue (2023): Forthcoming, Available for Pre-Order
Volume 14: 4 Issues (2022): 1 Released, 3 Forthcoming
Volume 13: 4 Issues (2021)
Volume 12: 4 Issues (2020)
Volume 11: 4 Issues (2019)
Volume 10: 4 Issues (2018)
Volume 9: 4 Issues (2017)
Volume 8: 4 Issues (2016)
Volume 7: 4 Issues (2015)
Volume 6: 4 Issues (2014)
Volume 5: 4 Issues (2013)
Volume 4: 4 Issues (2012)
Volume 3: 4 Issues (2011)
Volume 2: 4 Issues (2010)
Volume 1: 4 Issues (2009)
View Complete Journal Contents Listing