Research On QoS And Collaborative Filtering Web Service Recommendation Method

With the development of Internet technology, Web services recommendation and selection havebecome an important research in academic and industrial area, and quality of service (QoS) iscritical to recommend successfully. However, the values of QoS are likely affected by many factorssuch as server overload and network conditions when the service is running. Due to the dynamicenvironment of web service, the current existing service selection can not consider the inherentuncertainty effectively and the large deviation between the recommended results and the actualresults comes up. To solve the poor reliability of service selection caused by the dynamic nature andinherent uncertainty ignored, this paper proposes an improved web services recommendationmethod based on collaborative filtering, the introduction of this method make it possible that justanalysing the history information of QoS of web service to find out suitable and optimal servicewithout calling a web service.The collaborative filtering recommendation algorithm this paper poposed differs fromtraditional collaborative filtering algorithm. In terms of reliability of service, the inverse cloudalgorithm is introduced. In terms of similarity calculation, considering the personalized features ofweb services while calculating the similarity between users, and taking into account personalizationfeatures of users while calculating the similarity between the services. In prediction of defaultvalues of QoS, in order to alleviate the impact on the prediction performance caused by negative,this paper improve the traditional QoS prediction algorithm based on services and users, when theQoS predicted value for the target user is negative, the arithmetic mean of the QoS values ofservices or users are calculated to replace the negative. Finally, combining QoS predictionalgorithm based on services and users with the adaptive equalization weights to give the final result.To verify the efficiency of the algorithm proposed in this paper, large-scale data sets about QoSwhich contain1,500,000web service call records in real-world environment are used to simulate.The experimental results show the efficiency of the proposed algorithm.