基于QoS的Web服务组合Pareto推优研究

发布时间：2018-04-19 05:18

本文选题：Web服务组合 + 全局QoS优化　；参考：《南京财经大学》2014年硕士论文

【摘要】：Web服务近年以来一直处于高速发展阶段,能够提供相同或者相似功能的Web服务数量也越来越多,因而关于Web服务组合QoS的相关研究也越来越多。基于QoS的Web服务组合问题是典型的NP难题,当Web服务组合的候选服务数量规模较大时,如何进行服务组合的选择已经变得尤为复杂。本文针对现有解决Web服务组合QoS优化问题的不足之处,主要从以下几个方面做了相应的工作。首先,分析了以往求解Web服务组合QoS优化问题时存在的不足。大多数方法在处理此类问题时,往往将各个QoS按照一定的惯性权叠加或通过效应函数将多目标优化问题转换成为一个单目标优化问题,这些方法虽然在一定程度上解决了多个目标无法协调的问题,但是求解结果并不能客观反映用户的真实需要。所以本文在解决Web服务组合QoS优化上,考虑了将多个目标同时作为一个目标向量来考虑,将每个目标的具体的量值作为评价结果考虑。由于多目标优化问题的本质特征,多个目标互相之间往往很难同时达到最优,因而多目标优化问题的处理方式不再像单目标优化问题那样去简单的求取最优解。为此引入了Pareto解集的概念,将多目标优化求解过程与Pareto支配结合起来,使得求得结果具有很强的Pareto支配性,并将这些支配性很强的解作为优化的结果。其次,针对特定的Web服务组合QoS优化问题,通过对标准粒子群算法的分析,在其基础上进行了改进。由于标准粒子群算法主要适用于求解连续空间上的优化问题,而Web服务组合QoS优化问题是典型的离散型空间求解问题,所以将标准粒子群算法改进成为离散型粒子群算法,将原来粒子的飞行变为相应的跳动方式。针对粒子群算法的容易陷入局部收敛的特性,将遗传算法的变异策略引入到新的公式中,同时将蚁群算法的信息素的思想应用的新的粒子群公式中,从而使得新的粒子群算法兼具上述两者算法的优点。最后以一个典型的Web服务组合模型中的顺序结构模型为例,利用本文提出的新的粒子群算法,采用Pareto推优的方式,通过仿真实验来求解Web服务组合QoS多目标优化问题。对实验结果进行了相关分析,得到Web服务组合QoS优化问题求解速度的主要影响因素,并且通过与已有的方法对比,验证了本文方法的可行性和有效性,突出了本文方法在一定情况下求解过程中效率上的优势。从而为Web服务组合QoS优化问题提供了一种可行的求解方式。
[Abstract]:Web services have been developing at a high speed in recent years, and more and more Web services can provide the same or similar functions. Therefore, there are more and more researches on QoS composition of Web services.The Web service composition problem based on QoS is a typical NP problem. When the number of candidate services for Web service composition is large, how to select service composition has become more and more complicated.In order to solve the problem of Web service composition QoS optimization, this paper does some work in the following aspects.Firstly, the shortcomings of solving Web service composition QoS optimization problems are analyzed.When dealing with this kind of problem, most methods often superposition each QoS according to a certain inertia weight or convert the multi-objective optimization problem into a single-objective optimization problem through the effect function.Although these methods to some extent solve the problem of multiple objectives can not be coordinated, but the results of the solution can not objectively reflect the real needs of users.In order to solve the problem of QoS optimization of Web services composition, this paper considers multiple objectives as an objective vector at the same time, and takes the specific value of each goal as the evaluation result.Because of the essential characteristics of multi-objective optimization problems, it is difficult for multiple objectives to achieve optimization simultaneously, so the multi-objective optimization problem is no longer as simple as the single-objective optimization problem to obtain the optimal solution.In this paper, the concept of Pareto solution set is introduced, the multi-objective optimization solution process is combined with Pareto domination, so that the results have strong Pareto dominance, and these strong dominating solutions are regarded as the results of optimization.Secondly, based on the analysis of the standard particle swarm optimization (Swarm Optimization) algorithm, an improved Web service composition QoS optimization problem is proposed.Because the standard particle swarm optimization algorithm is mainly suitable for solving the optimization problem in continuous space and the Web service composition QoS optimization problem is a typical discrete space optimization problem, the standard particle swarm optimization algorithm is improved to discrete particle swarm optimization algorithm.Turn the original particle's flight into a corresponding pulsation mode.Aiming at the local convergence of particle swarm optimization algorithm, the mutation strategy of genetic algorithm is introduced into the new formula, and the pheromone of ant colony algorithm is applied to the new particle swarm optimization formula.Therefore, the new particle swarm optimization algorithm has the advantages of the above two algorithms.Finally, taking the sequential structure model of a typical Web service composition model as an example, using the new particle swarm optimization algorithm proposed in this paper, the multi-objective optimization problem of Web service composition QoS is solved by using the Pareto optimization method and the simulation experiment.Through the correlation analysis of the experimental results, the main factors influencing the solution speed of the Web service composition QoS optimization problem are obtained, and the feasibility and effectiveness of this method are verified by comparing with the existing methods.The advantages of the proposed method in solving the problem are highlighted.It provides a feasible solution for Web service composition QoS optimization problem.
【学位授予单位】：南京财经大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】