Service Pattern Based Service Composition Optimization Methods To Meet Massive Requirements

With the development of information technology, traditional services graduallytransform into modern services. Modern services assume as domain-specific andcomplicated multi-party interactive processes, which are supported by advanced ITtechnology and hybrided with realistic ingredients. The profound revolution of modernservice industry has led to significant changes in the goal of services, i.e., from pursuingthe basic functionality and performance of services to focusing on the overall servicevalue characterized by service profit and customer satisfaction; and from pursuingsimple service value to seeking for integrated service value-added. The core issuetowards enabling this process is to conduct value-optimized service compositionsaccording to customers’ personalized requirements. However, under the trend ofservicitization and massive integration characterized by cloud computing and big data,the building approach of domain services become more complicated, which entitles theoptimization problem with new features such as the massiveness of requests andcandidate services, and constrained service capacity. Those features pose newchallenges to service composition optimization with respect to problem scale,optimization criteria and related constraints.In the view of the defficiencies of existing methods in dealing with above problems,this paper presents a service pattern-based framework and related methods for solvingthe service composition optimization problem under domain environments with massiverequirements, with reference to the trend of requirement-service convergence exibitedby service engineering in solving domain problems. The main contribution of this thesisincludes the following aspects:(1) Systematic service composition optimization framework and related concepts.Considering the relevance among different optimization levels and techniques, thispaper presents a systematic framework and related concepts for solving the servicecomposition optimization problem under domain environments with massiverequirements. In particular, the features of massive requirements are analyzied andcorresponding measures are provided. And then a service pattern-based optimizationframework is proposed, which divides the optimization process into three stages, i.e.,massive requiremetns-oriented service configuration, service pattern-based solutionspace reduction and service pattern-oriented composite service selection. The mainproblems of each stages are identified, and the overall optimization process is described.(2) Massive requirements-oriented service configuration optimization. In the viewthat current research only focuss on single requriements, this paper extends the researchscope to scenario of massive requirements, and deal with the scenario by transformingthe problem specific to massive temporal sequential and concurrent requirements into that specific to groups of similar requirements. In particular, considering thespa-temporal relevance among the temporal sequential and concurrent massiverequirements, the requirements are segmented and clustered, so that they can share someoptimization results and the optimization efficiency can be improved. Correspondingmethods for reserving and distributing service capability are proposed so as to ensurethe equal distribution of limited service capability in both spatial and temporaldimensions. Considering differed emphases, scheduling-based and equivalence-basedservice capability distribution methods are provided, respectively.(3) Service pattern-based solution space optimization for service composition. Thisstage utilize domain experiences in business and optimization process to reduce thesolution space of service composition, so as to simplify the optimization problem. Inparticular, the concept of service pattern is proposed to describe common processes indomain business. Both service patterns and related advantageous services arepre-computed and organized, so that they can be reused in optimization. To adapt todynamicity in available services, a method is proposed for updating the advantageousservices on-demand. To further improve optimization efficiency, a Bayesian-basedcandidate service reductionmethod is proposed to empiriclaly and probabilisticallyreduce problem scale.(4) Service pattern-oriented composite service selection optimization. This stagefurther transform the optimization problems into single requirement-specific ones andsolve them based on previous stages’ outcomes, A candiate services filtering-basedmethod is proposed first to enable diversified and personalized solutions for differentrequirements within a same cluster. And then, different service selection strategies andmethods are proposed suitable for different scenarios, including uncertainty-basedservice selection strategy for key patterns, price-heuristic method for profit optimization,and improved artificial bee colony approach for dealing with massive candidateservices.Finally, to verify the theretical research results, a service composition optimizationtool is designed. Based on case study on the procurement business of the Smart HomeServices (SHS), the proposed theories and methods are verified.