Improvement Of Binary Particle Swarm Optimization And Its Application

Particle Swarm Optimization Algorithm is an algorithm that simulates an evolution mechanism of Natural biological. The conception is not only easy but also easily realized. It is without too many parameters to adjust.so it is international accepted soon and applied in many field. The algorithm can also deal with some nonlinear problems that are hardly solved by traditional search method. Basic PSO can not ideally deal with discrete problems, so this paper studies the binary Particle Swarm Optimization. It is better than basic PSO when dealing with discrete problems. Knapsack problem is a relatively common NP problem. All types of knapsack problem can be converted to0-1knapsack problem, so this paper focuses on the basic0-1knapsack problem as the study.To overcome the shortcomings that its convergence speed is slow and it is Easy to fall into a local optimum value, firstly this paper studies the Binary Particle Swarm Optimization Algorithm with greedy operator. The algorithm can reduce blindness of the searching process and improve the probability of the optimal searching. Then through the adjustment of parameters in the iterative process, this paper further proposed the improved greedy binary Particle Swarm Optimization Algorithm (IGBPSO).It can not only get better convergence effect but also improve the probability of the optimal searching. Moreover, through the experiment of using IGBPSO to solve the0-1knapsack problem, the algorithm will be tested and verified in terms of the probability of the optimal searching.Today is an era of information, this is a very high demand on our computing power, cloud computing came into being in this environment. Google proposed MapReduce model based on cloud platform at the Beginning of the21st century. It is a programming model for the mass data storage and data calculation under the large cluster. Combining the Binary Particle Swarm Optimization Algorithm with MapReduce model, this paper proposes an algorithm named MapReduce for Binary Particle Swarm Optimization (MRBPSO). Then we use parallel programs of MRBPSO to solve0-1knapsack problem. Experimental results show the feasibility and effectiveness when using MRBPSO to process Large-scale data.