QoS-Aware Web Service Recommendation Using Collaborative Filtering With Graph

Web service recommendation plays an important role in building reliable service-oriented systems for both the service providers and the active users.However,due to the diversity of users' requirements as well as the proliferation of similar web ser-vices in the Internet,traditional service recommendation is hard to accurately provide customized services to active users.In this paper,we propose a novel web service recommender model using collabo-rative filtering to improve the prediction of Quality-of-Services.Along with the online predicting part,the offline training part is added to overcome the sparsity of the user-item matrix.Benefiting from the accuracy of hybrid recommenders,we extend the idea of optimized predicting order and design the PTree and PGraph to describe the neighborhood.Furthermore,optimized strategy is introduced based on the PTree to promote the training performance,in which the iterative scheme is structure into the forest.Ad-ditionally,a new algorithm using adjusted topological sorting for PGraph is proposed to generate the optimized order while predicting.Both PTree and PGraph are generic enough to be applicable to general recommender systems.Finally,we conduct a series of extensive experiments to evaluate our proposed model,in which a real data set with 1.5 million invocation information is taken as input.The experiment results show that our model achieves higher prediction accuracy than other models.