| The swarm intelligence optimization algorithm mainly makes the algorithm more effective in problem optimization through collective collaborative behavior.All individuals collaborate within the search domain so that the entirety can move toward better and better areas.Brainstorming optimization(BSO)algorithm is the first swarm intelligence optimization algorithm with origin from modeling human activities at present.This algorithm takes advantage of human intelligence to deal with optimization problems.It is a very promising algorithm with potential capability far beyond existing classical swarm intelligence optimization algorithms.However,the basic BSO often shows the same drawbacks as those of traditional swarm intelligence algorithms,such as slow convergence rate,low convergence efficiency,and easy to fall into local optimum and so on.This paper mainly focus on improvement of the basic BSO algorithm through introduction of a so called "competition mechanism".First we define two characteristics of a community: diversity and evenness.By selecting the best group of seed solutions in term of diversity and evenness,we can speed up the BSO program significantly.The setting of BSO’s main parameters is also briefly discussed to avoid misguiding during its execution.Matlab simulation tests have been conducted about 10 typical benchmark functions,and results show that our improved competition-based brainstorming optimization algorithm is superior to the original BSO algorithm in terms of both the convergence speed and global balance.Fuzzy classification system is very useful in the field of knowledge representation and data classification.Fuzzy classification systems can generally be constructed by domain experts based on experience,but in most cases this experience is not complete,and the relevant data is relatively easy to obtain,so how to automatically construct a fuzzy classification system from the data has become a research hot spot in recent years.Because of its flexibility and resolvability,the fuzzy classification system can eliminate some ambiguity or uncertainty,making it applicable in many fields such as image processing,character recognition,text classification.Although the fuzzy classification system has good applicability,it still has many difficulties in the design process,such as the optimization of the rule base and the design of membership functions.This paper will use the improved competition-based BSO algorithm to construct a fuzzy classification system for the fuzzy rule generation and optimization process;and use a simple uniform distribution method to design the membership function;then use the fuzzy scores generated by the competition-based BSO fuzzy classification system to classify data samples.Simulation results show that in all different percentages of training data,compared with the existing AGFS in terms of sensitivity,specific index,and accuracy,the competition-based BSO fuzzy classification system performs better. |
