Abstract
The firefly algorithm is a nature-inspired metaheuristic optimization algorithm that has become an important tool for solving most of the toughest optimization problems in almost all areas of global optimization and engineering practices. However, as with other metaheuristic algorithms, the performance of the firefly algorithm depends on adequate parameter tuning. In addition, its diversification as a global metaheuristic can lead to reduced speed, as well as an associated decrease in the rate of convergence when applied to solve problems with large number of variables such as data clustering problems. Clustering is an unsupervised data analysis technique used for identifying homogeneous groups of objects based on the values of their attributes. To mitigate the aforementioned drawbacks, an improved firefly algorithm is hybridized with the well-known particle swarm optimization algorithm to solve automatic data clustering problems. To investigate the performance of the proposed hybrid algorithm, it is compared with four popular metaheuristic methods from literature using twelve standard datasets from the UCI Machine Learning Repository and the two moons dataset. The extensive computational experiments and results analysis carried out shows that the proposed algorithm not only achieves superior performance over the standard firefly and particle swarm optimization algorithms, but also exhibits high level of stability and can be efficiently utilized to solve other clustering problems with high dimensionality.
Highlights
Data clustering is an unsupervised learning task that involves the classification or grouping of data objects in a distinctive manner such that the items in one group or cluster are well defined and different from objects in other clusters [39]
We have proposed and implemented an automatic clustering algorithm based on the firefly algorithm (FA) and particle swarm optimization (PSO) algorithms
The simulation results have demonstrated that the hybrid FAPSO outperformed the two state-of-the-art clustering algorithms, namely FA and PSO, and other existing metaheuristic algorithms, namely, automatic clustering differential evolution (DE) (ACDE), GCUK, dynamic clustering particle swarm optimization (DCPSO) and DE, in a statistically meaningful way over most of the benchmark datasets used for the experiments
Summary
Data clustering is an unsupervised learning task that involves the classification or grouping of data objects in a distinctive manner such that the items in one group or cluster are well defined and different from objects in other clusters [39]. As discussed in Abraham et al [54], such an optimization criterion involves minimizing some measure of dissimilarity among data points within each cluster, while maximizing the dissimilarity among different clusters Since these techniques were first established, a number of different clustering methods have subsequently been proposed, namely, k-means, fuzzy c-means and simulated annealing, to solve clustering problems. A hybrid of two algorithms, particle swarm optimization (PSO) and firefly, referred to in this paper as FAPSO, is proposed with the main aim of overcoming the aforementioned limitations of the firefly algorithm, such as premature convergence toward local minima, due to the random initialization of cluster centres.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
