Research On Some Key Issues Of Dynamic Service Composition

With network technology's development, dynamic services composition technology is becoming one focus in research. Based on Semantic Web, Dynamic Service Composition technology requires that the network have semantic function, caple of discovering, selecting and composing services dynamically. Nevertheless, the research on how to enrich the web's semantics emphasizes on ontology, such as building relevant domain ontology for domain knowledge, building service ontology for Web service based on wide applications of SOA, etc. For this reason, the research on "ontology" is of vital importance for services composition technology.In addition, the requirement for individuation makes users' needs be increasingly regarded in the process of services composition, and providing users with such services that meet their individualized experiences is becoming the objective of the provider. Thus users' preference must be taken into consideration if more individualized and satisfying services are to be composed in the services compositon.Aiming at solving the above problems, this paper achieves the following results:1) One services composition method based on attribute semantic association is advanced. This method converts the input and output parameters of the service into its relavant attributes, and builds the rule for attribute semantic association to make services compostion. Being able to reason based on ontology by using engine, and eliminating the complicated mediation between various service interfaces by using complex mechanism, this method promotes the efficiency and feasibility of the services composition, which makes it a light weight services composition method. Additionally, the domain services ontology that has been built can effectively organize services according to the domain classification concept of the ontology, differentiate the functions between various services, and reduce the searching space in services selection.2) In order to apply preference description to the dynamic services composition, a method named PODDA is put forward. This method adds semantics to the service preference of the user and decomposed this preference into many sub preference sets or attribute preference sets. By abstracting sub preference sets from the relations of the domain ontology's concepts and services' attribute preference sets from the relations between the concepts, the building of the relevant ontology could better utilize the existing resource in domain ontology and express the user's preference under multiple restrictions more profoundly. In the end, a preference ontology on tourism is built to test the feasibility and rationality of PODDA.3) On the basis of the above research on preference, we make an analysis of a service recommendation issue with a special feature—the law of Diminishing Marginal Utility. By using the theory of consumer's action in economics, and such concepts as preference and utility, as well as the law of Diminishing Marginal Utility, we make respective researches on single service recommendation and on multiple services composition recommendation. This method can theoretically maximize the utility of the composed services on the basis of a certain budget.In the individualized service composition method mentioned above, we need to select in a great amount of services such ones that meet users' demands, which decreases the efficiency of services composition and make it urgent to find effective method to optimize service composition. Accordingly, this research puts forward two service selection and services composition optimization algorithms: one is for users' QoS attribute constraint, and the other introduces users' preference ontology and adds preference weight to make the optimized result more fit for the users' individualized needs.4) In order to optimize services composition and to be adapted to the dynamic and instable web services, as well as the limitation of multi-QoS attributes in the process of services composition, we put forward an algorithm named Multi-pheromone and Dynamic Update Ant Colony Optimization Algorithm (MPDACO). This algorithm can fit for the service futility, service increase and QoS change in the process of services composition optimization. In addition, the algorithm is improved to accelerate convergence to optimized result based on experiment. In the end, we also put forward a method that is based on ACO and can find optimized result heuristically, to be applied to increase service selection efficiency when the volume is great.5) In order to address the challenges from services selecting in great volume, the information reuse in services selection, and the individuation of the services composition, a method of services selection applying Ant Colony Optimization (MSSACO) is put forward. The MSSACO sets multiple pheromones and pheromone weights to note the preference to different properties of a service. Having a unified services selecting model that is based on a comprehensive analysis of different services composition methods, the MSSACO can make individualized services selection according to the record of a service, and can be adapted to the dynamic change of services composition process and QoS. Making some improvements on the strategies of ACO, the MSSACO can find if the optimization is converging to local optimized value, in which case can take measures to change its direction, and as a result reduce the probability of the algorithm to converge to local optimized value.