CAJ | 학술논문

为了解决Pareto遗传算法在每一次进化操作中都要构造当前进化群体最优解集而影响运行效率的问题,提出了一种面向Parfto最优遗传算法的服务组合方法,以实现Web服务组合的全局优化.用伪二叉树法则构造目标函数的Pareto最优集合,再进行Pareto最优解集排序,最后采用个体相似度计算来确定遗传算法的适应度函数,由此获得一组满足约束条件的Pareto最优解服务集合.实验验证表明,所提方法可以提高多目标遗传算法处理服务组合效率的问题,即使在服务规模较大的情况下,所获得的解与最优值的比率仍能接近90%的水平.
위료해결Pareto유전산법재매일차진화조작중도요구조당전진화군체최우해집이영향운행효솔적문제,제출료일충면향Parfto최우유전산법적복무조합방법,이실현Web복무조합적전국우화.용위이차수법칙구조목표함수적Pareto최우집합,재진행Pareto최우해집배서,최후채용개체상사도계산래학정유전산법적괄응도함수,유차획득일조만족약속조건적Pareto최우해복무집합.실험험증표명,소제방법가이제고다목표유전산법처리복무조합효솔적문제,즉사재복무규모교대적정황하,소획득적해여최우치적비솔잉능접근90%적수평.
A new method of service selection is proposed to implement the global optimal selection of Web services,and to avoid the routine in a genetic algorithm where an optimal evolving set is built for every current evolving operation to impose a negative influence on the efficiency of the algorithm.The pseudo binary tree'S rule is used to produce a set of optimal Pareto solutions. Then the optimal Pareto solutions are sequenced and individuals'similarity of each Pareto solu-tion is calculated to determine the fitness function in the genetic algorithm.Experimental results show of proposed method generates a result that has a similarity up to 90% compared with the optimal result,even in the case of a large number of services.