Online Recommendation Approach For Web Service Based On Reputation And Network Map

With the popularity of service-oriented computing,more and more Web services with the same or similar functionality are emerging.When the number of Web services is too large,it becomes very difficult for users to choose the best Web service.Therefore,how to help users learn more about Web services and choose the Web services has became an important issue in current research.Existing research shows that users pay more attention to the quality of services(QoS)when meeting functional requirements.At present,most researches on Web service selection,discovery and recommendation are based on the prediction of QoS values,improve the recommendation quality by improving the accuracy of QoS value prediction.The analysis of historical data shows that the user's network environment is an important factor affecting the QoS value of Web services,but there are few academic research in this direction.However,scholars haven't enough attention to the untrusted users in the area.The false data submitted by these users has a great negative impact on the prediction of the QoS value,and reducing the performance of the recommendation system.In addition,the Web service is in a running state,and the QoS value of the service may change at any time.The method of using off-line learning technology to train QoS prediction model is unable to deal with the real-time update of data and the computation of large-scale data,which cannot meet the real-time performance of Web service recommendation.Aiming at the above problems,this paper proposes an online Web service recommendation method based on reputation and network map.This method mainly improves the recommendation quality of Web service recommendation system from three aspects:(1)Based on the user's network information,the network map is built to measure the network distance between users and divide the user community according to the network distance.Belonging to the same user community the service users against distance between differences exist problems,put forward the target users,the optimized plan for the different neighbor assigning different weights to improve QoS prediction accuracy.(2)We propose a QoS prediction method based on the reputation and network map,according to the user which is the same community submits the QoS data calculate the user's reputation,reduce impact on the QoS prediction accuracy.The reliability of QoS prediction model is improved.(3)An online updating method for the dynamic change of QoS data is presented,the collected QoS data flow is preprocessed,and the normalized QoS data is combined with some historical data to build a new user-service matrix.The online learning technique is used to train the matrix decomposition model to improve the computational efficiency and ensure the real-time performance of the prediction.Build a Web service recommendation list based on QoS data in real time and recommend appropriate Web services to users.In this paper,the feasibility of the algorithm is verified by experiment and result analysis.In addition,compared with the traditional recommendation algorithm,the proposed algorithm improves the recommendation quality of Web services to a certain extent,while ensuring the real-time performance of the recommendation system.