Context-aware Web Services Recommendation For Modeling Spatial Correlations And Weighted Rating Effect

In recent years,more and more Web services have been published on the Internet,which results in a large quantity of Web services with the similar functionality.Thus,how to recommend a personalized Web service to a specific user becomes gradually an important and challenging research direction in services computing.Traditional collaborative filtering methods could not satisfy with the emerging requirement of Web services recommendation.As a result,context-aware Web services recommendation(CASR)methods become more and more popular.However,existing CASR methods have two drawbacks.First,CASR methods mostly use the temporal or spatial context of users or Web services to find similar neighbors for the target user or the target Web service.However,they didn't fully consider the correlations between users' context and Web services' context,which might have an important effect on users' preference to a Web service.Thus,service recommendation results are difficult to respond to dynamic changes in user or service contexts.Second,when computing the similarity using user-service QoS matrix,existing CASR methods neglect the weighted rating effect and consider every QoS value equally.As a result,it is difficult for the user to provide a significant reflection of the user preference to the Web service.This paper proposes a context-aware Web services recommendation method by modeling both spatial correlations and weighted rating effect(CASR-SCWRE).The method is intended to provide users with a personalized Web service recommendation mechanism.On one hand,the proposed algorithm could explore the effectiveness of spatial correlations on user preference expansion.On the other hand,the proposed method could significantly reflect the significant impact of different QoS values on user similarity or service similarity.The main contents of this paper include: First,by taking into account the dynamic characteristics of geographical location for both the user and the Web service,we model the effect of spatial correlations on user preference expansion by providing the personalized filtering of the services before the similarity computation.Second,we propose an enhanced temporal decay model in similarity computation,which incorporates the weighted rating effect into the traditional temporal decay model to improve the prediction accuracy.Third,we use the invocation records of Web services filtered from the above two steps to make QoS predictions using Bayesian inference,and perform the QoS prediction by using Bayesian.Finally,this paper conducted a set of comprehensive experiments based on a real-world Web service dataset WS-Dream.The experimental results demonstrate that the proposed method significantly outperforms existing Web service recommendation approaches.