The Research On The QoS Evalutaion And Recommendation Based On Service-Side Feedback

This paper is supported by the National High-Tech Research Development Program of China (863 program) under Grant No.2007AA01Z139, and research on service quality evaluation and recommendation.In recent years, with the rapid development of web services, the number of services on the Internet shows a dramatic increase, which results in the increase of the number of services which provide the same functions. Researchers have generally realized that the non-functional attributes related to the quality of services become more and more critical to the success of service-oriented application.Aiming at these issues, the main objective of this paper is to establish a service recommendation system based on Qos attributes. The system ranks the services evaluation according to the quality attributes of services. This paper proposes a scalable web service quality framework. In the situation of new service qualities are being increasingly proposed, the framework is flexible to support the extension of service quality attributes, which enable the easiness of description and measurement of the new quality attributes at service-side. In the basis of this framework, the quality information is collected in real-time by service-side and is feed backed to registration center.Secondly, the service recommendation system designed by this paper is used to enhance the registration center. In the past researches, the selection of services is simply based on the ranking of attributes. This paper presented an optimized neural network model to analyze the collected Qos information in real-time. The information of services which were called during the past time is used as training data. With the accumulation of information of services called by users, the training process proceeds. Also the users'preferences are analyzed based on data-mining. Considering both the users'preference and quality of services, the appropriate service can be provided to the clients.Thirdly, in the key issues of the neural network model-the design of the hidden layer, PCA is adopted to adapt to the extension of quality attributes. With the assistance of PCA, the system can redesign the neural network automatically after the extension of the inputs of neural network. By combining the neural network and PCA, the fully automatism of web services can be implemented.In this paper, we set up a service quality evaluation and recommendation model based on neural network which can evaluate service automatically and objectively, this model laid a good foundation for SOA theoretical research.