Dynamic Particle Swarm Optimization Algorithm And Application

With the development of computer and communication technology, the data stream as a new data form, whose feature is super-speed, succession and magnanimity, is subjected to more and more concern. In order to make real-time analysis and explore the dynamic changes of data stream, a data stream clustering technology, which is different from the traditional cluster technology, is needed. And evolutionary algorithm has been applied into cluster analysis and it can deal with traditional cluster technology as a optimization problem. As a representative of the evolutionary algorithm, particle swarm optimization algorithm has been widely used to static and dynamic optimization problem, due to its significant advantages of simple principle, fast convergence speed, good robustness, etc. The people treats the data stream clustering problem as a dynamic optimization problem and applies a new dynamic particle swarm optimization algorithm to process it. The main work of this paper is as follows:1. A improved of dynamic multi- swarm particle swarm optimization with orthogonal learning(OLMPSO) is proposed. Firstly, a parent swarm is ultilized to explore the search space and some child swarms to exploit promising areas found by the parent swarm. An orthogonal learning(O L) strategy utilizing previous search information(experience) more efficiently is applied in these child swarms to improve the convergence speed. And when the search areas of two child swarms overlap, the worse child swarms will be removed. Moreover, in order to quickly track the changes in the environment, all particles in a child swarm perform a random local search around the best position found by the child swarm after a change in the environment is detected. The experimental results show that the efficiency of O LMPSO for locating and tracking multiple optima in dynamic environments is outstanding in comparison with other particle swarm optimization models.2. A new data stream clustering algorithm based on a improved dynamic multi-swarm particle swarm optimization with orthogonal learning(O LMPSO-Stream) is proposed. The algorithm aims to build a the optimization model in order to transform the data stream clustering into a dynamic optimization p roblem, which regards the process of data changes over time in data stream clustering as the changing environment in dynamic optimization. And FCM algorithm improved with the operator of particle swarm optimization in O LMPSO is applied to speed up the clustering speed and improve the quality of clustering. O n this basis, an environmental prediction mechanism is proposed to adjust the particles accordingly after the environment changes. Finally, the experimental results show that the efficiency of O LMPSO-Stream is outstanding in comparison with other data stream clustering algorithms.