A Kind of Diminishing Step Fruit Fly Optimization Algorithm

Abstract:

Parameters of Support Vector Machine are playing an important part in learning performance and generalization capability. The randomness and blindness in selecting SVM model parameters artificially could be eliminated by using group intelligent optimizing algorithm. FOA is a kind of group intelligent optimization algorithm. It has some advantages such as global convergence, connotative parallelism and fast operating speed. However, its optimum efficiency is very sensitive to the length of fixed step. In the course of optimizing, if the step is oversize, it will have preferable global optimizing performance and weak local optimizing capability. On the contrary, if the step is undersize, the local optimizing capability would be powerful and it will have the most probability to lapse into local extreme value. Therefore, a kind of algorithm named Diminishing Step FOA is proposed, the step length minishes progressively along with the process of foraging. So that it would have preferable global optimizing capability in early stage and preferable local optimizing capability in later period. And then, a dynamic balance will be achieved between global and local optimizing capability. The experimental results show that the SVM model using DS-FOA has optimal forecast precision and effect.

• Periodical: Applied Mechanics and Materials (Volume 487)
• Pages: 687-691
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.487.687
• Online since: January 2014
• Authors: Jun Yan Hou*, Bing Wang
• Keywords: Diminishing Step, Fruit Fly, Parameter Optimization, Support Vector Machine (SVM)

References

[1] Cortes C, Vapnik V N. Supporter vector networks. Machine Learning, 1995, 20(3): 273-297.

[2] Chapelle O, Vapnik V, Bousquet O, et al. Choosing Multiple Parameters for Support Vector Machines. Machine Learning. 2002, 46 (1) : 131-159.

[3] Zhu Fengming, Fan Minglong. Chaos Particle Swarm Optimization Algorithm for Optimizing the Parameter of SVM. Computer Simulation, 2010, 27(11): 183-186.

[4] Dai Dongxue. PSO algorithm and application based on chaos thought. Journal of Middle China science and technology university, 2005, 33(10): 53 -55.

[5] Sun Ya. BP neural network face identification algorithm based on PSO. Computer Simulation, 2008, 25(8): 201-204.

[6] Wen-Tsao Pan. A new Fruit Fly Optimization Algorithm: Taking the financial distress model as an example. Knowledge-Based Systems 26(2012): 69-74.

DOI: 10.1016/j.knosys.2011.07.001

[7] Data source: http: /www. gw. com. cn.