Research On Web Service QoS Prediction Based On Location-nearest Neighbor And Matrix Factorization

With the continuous development of Internet technology, Web services as a best practice model of service-oriented architecture has being used to provide new network service, and the new service-oriented Internet era is approaching. Under this environment, the number of service is growing fast, massive services are deployed in a variety of cloud center and network environments, and traditional service discovery and selection methods could not meet the needs of users. Thus, the service recommendation technique has been developed. Service recommendation technology achieves the transformation from the passive acceptance of user requests to the active perceivable change of user needs, and can freed user from the mass of the services by providing users with precise recommendations.Due to the instability of different networks and different users of the network environment, the same Web service for different users may have different QoS values. Moreover, letting all users call all of the candidate services to select the optimal Web service is impractical. Therefore, the Web service QoS prediction which can assist the service recommendation system recommended has become a serious problem.This paper mainly studies the Web service QoS prediction problem, summaries the main ideas of common methods and analyzes the existing problems in them. Since Qos of Web service has great relevance with the user’s position, we propose a Web service QoS prediction method based on location-nearest neighbor and matrix factorization. Specifically, the main contents and contributions of this paper are as follows:First, we introduce the theory of Web service and the recommendation system, discuss the basic concepts of QoS, recommendation system classification and analyze the main issues of the current recommendation system.Secondly, we introduce the traditional matrix factorization techniques and similarity calculation algorithm. On this basis, we put forward the user Neighborhood model and service Neighborhood model based on location information. Following the election of neighborhood of users and services, we combine the neighborhood information of users and services to the matrix factorization model as manner regularization terms, by gradient descent algorithm to solving the minimum objective function to improve the prediction accuracy of QoS values.Finally, we use Python to program for the relevant experimental study, and compare with the current state-of-art algorithms to verify the validity of our method.