Research On Key Technologies Of Web Service Discovery And Selection

Web Service Composition technology is an effective way of service and resource integration which achieves a new and more complex business logic through the reuse of existing Web services.In the delivery process of the composite service schema, the workflow engine needs to search the collection of executable candidate Web services from the Internet based on the functional requirements of the abstract composite service schema,the reputation of these candidate services will have an important impact on the successful delivery of the composite service. For each abstract service of the composite service schema, there may exist a lot of Web services in the network with same or similar functionality and with different quality of service, it will form massive composition schemas to select appropriate Web services from the candidate service set and build executable composite service,namely the "composite explosion". This is a typical NP-hard problem, and its efficiency will have an important impact on the delivery qulity of the composite service. In the process of the concrete service execution of the composite service schema, it may cause fail due to the malicious service or the changes of service quality, so it needs to perform a failure recovery functionlity for the composite service, then it is very important for the quality assurance of composite services about how to select the effective and replaceable services.This paper carries out relative studies concerning the above problems, and proposes appropriate solutions for them, main research achievements include:(1) Aimed at the problem of that the low precision of service discovery will affect the successful delivery process of the composite service, we propose a trustworthy service discovery method called TSDMACS based on modified ant colony algorithm for the unstructured P2P network. This method adopts a set of novel policies to control the behavior of the ant colony, such as the dynamic ant colony policy, the sub-ant policy, and the reputation evaluation policy for service peers etc., service peers with high reputation change their neighbors and make self-organization under the recommendation of ants.Through the comparison with traditional MMAS algorithm, it shows that the proposed algorithm can ensure higher trustworthy service precision and better integrated service discovery performance in common complex network environment.(2) Aimed at the problem of massive composite service schemas,We propose a Web service selection method called MDPSO based on improved discrete particle swarm algorithm. The MDPSO algorithm is proposed based on sub-particle circular orbit and zero-value inertial weight, which is applied in web service composition optimization problem. MDPSO adopts trigonometric-function-based non-linear dynamic learning factors and a prediction method for population premature convergence, which can keep better balance between the local exploring ability and the global converging ability of particles. Experimental results show that MDPSO algorithm can achieve better performance than traditional PSO algorithms in WSC optimization. These concepts and methods provide a new thinking route for application researches of WSC problem. Some useful conclusions are obtained through the analysis and explanation of the experimental data, which lay a solid foundation for further researches.(3) Against the problem of fault service replacement in the composite service delivery process,we present a trustworthy Web service recommendation method called TWSRCF based on collaborative filtering. First, it searches a set of Web services with same or similar functions according to the user’s requirement. Second, it computes the collection of users that have similar preferences for the target user based on their common evaluations, and obtains the recommendable user set for each candidate Web service, then computes the recommendation degree for each service using its similarity, evaluation degree and trust degree, and sorts the candidate Web services according to the recommendation degree to the target user. The experimental results show that the referral effect increases with the increase in the number of user evaluations, and with the increase of malicious user evaluations in the system, compared to other service recommendation methods, it only has a slight impact on the referral result of this method.