Web Service Recommendation Approaches Based On Location Awareness And Collaborative Filtering

With the rapid growth of Web services,it is necessary to build an efficient Web service recommendation system.The key problem of recommending high quality services to users is how to obtain the QoS value of the Web service.Although the user can evaluate its QoS by invoking the Web service individually,it is unrealistic to evaluate the QoS of a large number of candidate services within short t ime because that users are usually not experts in evaluating the qualities of services.Given that the QoS of Web services is usually dependent on specific users,in recent years,many efficient methods using collaborative filtering to personalize the QoS prediction and service recommendations were proposed.Some of them have been evaluated with real-world Web service QoS dataset.However,the traditional collaborative filtering technology performs poorly when the data is sparse,and it also has cold start and poor scalability and other issues.In addition,due to the network latency and network conditions,users in the same region are more likely to observe similar QoS on the same Web service.This feature,however,was rarely considered in prevous work.To address the above issues,this paper presents a new Web service QoS prediction and recommendation method.The main contributions of this paper are as follows:(1)A location and clustering-based collaborative approach for Web service recommendation.Firstly,users are clustered according to the their locations by using the correlation between the service QoS and user location.Based on the clustering results,part of the missing QoS values can be filled.Then it finds the most Top-K similar users of the active user to predict the unknown QoS values of the target service.This approach can effectively solve the sparseness of Web services data and cold start problems,while achieving a better balance between the precision and coverage rate.To validate the accuracy of the proposed method,we have carried out a series of experiments on the real Web services data set.The experimental results demonstrate the effectiveness of the proposed method.(2)A QoS-aware Web service recommendation approach based on locat ion-aware factorization machine.The approach firstly identifies a similar neighborhood for each user(or Web service)according to its network location,and then uses a neighborhood-based regularization term to revamp the factorization machine model.Cons equently,the unknown QoS values of candidate Web services can be predicted for the active user,and high-quality Web services are thus recommended.Experiments conducted on real-world Web service data sets show that the approach outperforms the state-of-the-art collaborative filtering approaches regarding to the accuracy.In addition,the prediction time of the approach is almost linear to the data size,indicating that the data sparsity and scalability issues in large-scale Web service recommender systems can be addressed efficiently.