Research On Web Services Recommendation Algorithm Based On QoS

The number of Web services that can provide the same functionality is increasing now.It is difficult for users to find services that meet their needs using functionality.Quality of Services(QoS)has become an important criterion for users to choose Web services.How to predict users' feedback information on the other Web services they have not visited is a hot research topic.The recommendation system is a good way to predict QoS.Collaborative Filtering(CF)is widely used in recommendation system.Taking the user-based collaborative filtering as an example,the key step is to calculate the similarity between users by QoS data and find a similar user set to predict the QoS data.The paper mainly improves three aspects.Firstly,the traditional method only considers the correlation of two vectors which are composed of common data when calculating the similarity,while does not consider the length of the vector and inherent attributes of the user or project.The proposed method considers the length of the vector and it can cluster users or services by location information.Secondly,the traditional method assumes that the user feedback data is credible and does not take into account users with abnormal and malicious evaluation data.This article proposes a method to exclude untrusted data by user credibility information.Finally,a data perturbation privacy protection algorithm is improved in this paper by adding a personalized component.Combining with the first proposed algorithm in the paper,the protection algorithm can reduce the impact of data perturbation on similarity and meet the personalized privacy protection need.The average absolute error and the root mean square error are used to evaluate the prediction result.