Cross-iteration Particle Swarm Optimization And Its Application In Combination Auction

In recent years, with the rapid development of computer science and technologyand e-commerce, combination auction have great pontential in application prospects.How to get more effective solution of combinatorial auctions is the core issue incombination, which have direct effect on the applicaton prospects and practicality. Asthere is strong correlation among goods and bid can not be divisible, combinatorialauction can express correlation value among goods better, that is to say it can have moreand more effective way to express to real needs and preferences of the bidders. As resultof it, it can reduce the risk of the auction and increase the income of the auctioneer.Therefore combination auction is more effective way for resource alloction. Howeverthe solution space of combinatorial is huge, such as there is n goods, then it have2n1portfolio. As the n increases, the number of combinations bids is exponential growth.Combinationial auctions determined problem has been proved to be an NP-hardproblem，so it have good sense of reality to do research on combinatorial auction.In this paper, the cross-iteration particle swarm optimization and multigroup PSOhave been proposed and four benchmark functions are been used to verify itsperformance. Then discrete particle swarm algorithm strategy is studied. At last, discreteparticle swarm optimization is used to solve the combinatorial auction combining ofgreedy repair operator. Main research content of this paper includes the followingaspects:(1) The hybrid particle swarm optimization is proposed which is the cross-iterationof global particle swarm optimization and local particle swarm algorithm. Then fourbenchmark functions which include sphere, Rastrign, Rosenbrock and Griewank areused to vertify its performance in30,50and100dimensions. It has been proved that theconvergence of the cross-iteration particle swarm optimization is faster than the globalparticle swarm optimization and it has much higher convergence precision. Then themultigroup PSO is more outstanding in function optimization.(2) Through the advance study, the discrete particle swarm optimization is proposed,which maintain the same basic form of the particle and map the location of the particleto0or1. It has been showed that this DPSO is better than BPSO and proposedcross-iteration discrete particle swarm is the best on comprehensive.(3) OR bids and XOR bids models of combination auction is established and itssolution space are defused.On the basis of it, the greed repair operator is been used torepair the solution to make it to be a legitimate solution.(4) Finally, the CATS2.0which is the standards of the combinatorial auctions togenerate different scale data is been used to validate the effectiveness of the cross-iteration particle swarm optimization under various economic distribution.