Font Size: a A A

Research On Personalized Service Recommendation Approaches Based On Quality Constraint

Posted on:2020-06-10Degree:MasterType:Thesis
Country:ChinaCandidate:L WuFull Text:PDF
GTID:2428330575465328Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
With the rapid development of the Internet and the broad application of service-oriented architecture,a large number and variety of services have been published,e.g.,IoT services,web services and cloud services,etc.For a service,there will be a large number of candidate services similar to its functional attributes but different from its non-functional attributes,which makes it more difficult and important to recommend personalized services for users effectively.Therefore,service recommendation has become a very important research topic,and has strong practical significance.Existing web service recommendation approaches employ content-based,collaborative filtering,utility-based and skyline-based.Content-based recommendation is based on a large number of historical behavior data of users,mining user preference types,and then recommending services according to the preference types.The recommendation method based on collaborative filtering is to find the neighbor user set for the target user according to the user's historical behavior data,then predict the QoS value of the unknown service for the target user based on the set,and finally select top-k services to recommend to the user.Utility-based recommendation calculates the utility value of candidate services through some calculation method,and then makes recommendations according to the utility value.Skyline-based recommendation mainly uses skyline technology to select those services with the best global quality of service from a large number of candidate services to recommend to users.However,these methods ignore several key issues.First,how to find representative services based on the quality constraints(e.g.failure rate,delay,etc.)given by a user.Secondly,while looking for representative services,we also need to consider the correlation of users' quality constraints in different dimensions,because users' quality constraints may be vague and willing to make concessions in a certain dimension.Finally,in the actual environment,the QoS value of the service will change due to the dynamic environment and other factors,that is,the QoS value is uncertain.It is also a problem how to accurately recommend representative services to satisfy users' preferences in such an environment.In order to solve the above problems,this article proposes two service recommendation methods in two cases.In the case of certain QoS,diversified quality centric service recommendation method(DQCSR)is proposed.In the case of uncertain QoS,a recommendation method based on p-Dynamic Skyline and KNN(PDS-KNN)was proposed.The main contributions of this article are as follows:(1)Firstly,A diversified quality centric service recommendation method(DQCSR)is proposed.There are three stages in this method:firstly,Dynamic Skyline services are found from candidate services based on users'quality constraints,which ensures that the selected services are representative.Secondly,K-means algorithm is used to cluster the Dynamic Skyline services so that similar quality value services can be clustered together as much as possible to get k clusters.Finally,in order to select a service from each cluster,two selection methods are proposed,which are cluster-based and coverage-based,respectively.This process can ensure that the selected services are as diverse as possible.(2)Secondly,this article also considers how to find some representative services to satisfy users' preferences in the case of uncertain QoS.Therefore,this article proposes a recommendation method based on the mixture of p-Dynamic Skyline and KNN(PDS-KNN).The process of this method is as follows:Firstly,users need to give quality constraints sr and domination probability threshold p.Secondly,services in candidate service sets at each time are mapped to the original service space according to their quality values.Then,the p-Dynamic Skyline services are found by calculating the dynamic domination probability between services.Finally,KNN method is used to select the k services closest to the user quality constraint from the p-Dynamic Skyline services and recommend them to users.(3)Finally,extensive experiments are conducted on a real-world dataset to evaluate the effectiveness of our proposed approaches.In the case of certain QoS,the experiment is based on the QWS data set,which contains 2507 real-world Web service quality information.In the case of uncertain QoS,the experiment is based on QWS to randomly generate a normal distribution data set,which contains 2507 quality information of Web services at 24 times.
Keywords/Search Tags:Service Recommendation, Quality Constraints, Quality Correlation, Quality of Service (QoS), Dynamic Skyline
PDF Full Text Request
Related items