Web Services Selection Based On Fuzzy Neural Network

With the mature of the Web framework, W3C defined the Web servicesarchitecture. Web service is a modular, self-descripted, self-contained applicationcomponent which can be accessed through Internet. The Web service cycle includes:service advertisement, discovery, selection, combination and calling. Web servicescan also be arranged reasonably by the relation degree of the candidate servicesdiscovered by UDDI and the user service request, or some non-functional request,such as QoS. In fact, along with the rapid increase of the web service providers,services with same or similar functions are provided to users. So, the non-functionalattributes, such as QoS, has become the important considerations in service selectionand sequencing.Artificial Neural Network, the main branch of Computational Intelligence,simply simulates the human brain structures and the ways of thinking by parallelnetwork. Then, Fuzzy system simulates the decision process through the impreciseinformation based on the theory of fuzzy mathematics. It is an effective way toquantitatively reveal and analysis fuzzy.In this paper, attempt to form a web services selection algorithm based on theFuzzy Neural Network by the organic combination of Artificial Neural Network andfuzzy theories, and simulate it in Visual C++environment. Thus, we make full use ofthe real-time controlling advantages of the ANN, and join the membership function inthe three-layer structure. In the preparation stage, the teacher signal of the output layeris also improved by joining a mixed weight calculated from user subjective weightand the objective weight based on the real-time entropy information.This paper briefly introduces the research background and significance of webservices discovery based on QoS. Then the existing measure model and algorithmsrelated. The main content is the basic principle, hierarchical of FNN and the deduce ofthe core BP implementation process.We divide the neuronal connection weights into a linear synthesis of objectiveand subjective ones. The algorithm first reads the user preference information bynatural language from QoS agent; convert it to the fuzzy condition rules sentences;express the user request vector with triangle fuzzy number. After that, all the QoS values are built to a standardized matrix.