Research On User-based Service Selection Methods In Large-scale Environments

With the development of cloud computing technology as well as rapid popularization and promotion of distributed application systems such as Web services,a growing number of service resources with high-dimensional characteristic on the Internet offer more choices for service requesters. However, traditional service selection methods cannot help users select desired services quickly and effectively from large amount of candidate services sets any more. Moreover, Users get the experience of continuous innovation,which propose new requirements for the personalized service selection. To address the problems above,this thesis does research on the service selection methods in large-scale environments and proposes the user-based service selection methods, the main innovation points are as follows:From the perspective of elementary service selection,we propose a skyline service selection algorithm based on users’ preference in this thesis. A preference dominance relationship is firstly derived from the definition of desired service, which contributes to filtering skyline services set in accordance with users’ preference. Relative entropy is then introduced to calculate the distance between desired service and each skyline service. Top-k sorting results in skyline services set are finally selected, followed with a dynamic correction algorithm of user preference degree. Based on users’ choices of services, the algorithm can calculate the preference degree adjustment function and rapidly correct user preference degrees on different QoS attributes. Simulation experiments and results demonstrate that the proposed approaches can achieve an objective of an orderly skyline services set and differentiate potential preference differences on Qo S attributes, which have the performance of a higher user satisfaction and good extensibility.From the perspective of composite service selection, we propose an optimization algorithm for composite service solution space based on users feedback. First of all,users feedback information is integrated with the optimization of composite service solution space. Their positive and negative feedbacks are considered to calculate the generation and reduction of version space; Secondly, bayesian formula is used to calculate the posteriori probability of service hypothesis in version space,and the posteriori probability is used to measure the coincidence degree between service hypothesis in version space and user feedback; Finally,a set of services with the largest posterior probability is obtained. Since the generation and reduction of version space can be done offline, so the demands of large-scale service scenario for online optimization efficiency are met.Based on the theory above,this thesis designs a user-based service selection simulation prototype system and gives an application demonstration. The implementation of system follows the process of demand analysis,general design and module design in detail,and accomplishes Uses Preferences Management,Candidate Services Searching modules,which illustrates the feasibility and effectiveness of the methods proposed in this thesis under the dynamic scene.