Prediction Of Service Reliability Based On Group

Service reliability prediction has been an important research topic in the field of service computing,and its prediction results have the vital impact on service recommendation and selection.With the development of the Internet and the continuous extension of related technologies,the hardware and software environment of the network over Internet users presents complex,dynamic and heterogeneous characteristics,which bring challenges to the reliable implementation of the services provided by service providers.Therefore,in the field of service computing,how to select high reliability services for users in a large number of similar services has become a concern.At present,the reliability prediction method is mainly from the perspective of the user or the service itself,but neglects the influence of the network environment on the service reliability prediction.In addition,the user's group is different,and even if the experience with some Web services is basically similar,the feedback may vary considerably.Hence,the prediction of service reliability from the group can ensure the accuracy of service reliability prediction.First,from the perspective of the user group environment,this thesis mainly focuses on two aspects of the current service reliability prediction methods: ignoring the impact of the user's own environment on the service reliability and the lack of performance optimization of the service reliability prediction method.A collaborative prediction method with prior steps is proposed.In this method,a priori step is designed firstly,in which the user similarity threshold is calculated and the cost of prediction is reduced.Then,using multiple user environments in inherent group which users belong to,the environment change of individual user is simulated,and the service similarity is calculated by IPCC calculation method.Furthermore,based on the complete reliability information matrix,the UPCC method is used to calculate the user similarity based on the service.Finally,coordinating the service similarity and the user similarity,and the reliability prediction results are obtained.Simulation results show that the proposed service reliability prediction method has better prediction effect.Secondly,from the perspective of big data implementation,this thesis proposes an application model of service reliability prediction to the group under the big data.Aiming at the problems that service reliability prediction's big consumption much and it is difficult to be applied in real-time big data,a group-oriented application model of service reliability prediction is proposed,which includes user similarity solution in inherent group,dynamic computing similarity threshold,low cost service similarity reliability matrix method and off-line computing reliability matrix model,besides,distributed and calculated reliability matrix through CDN,based on geographical location processing user requests,user requests are guaranteed to get a quick response.Finally,based on the above methods and theories,this thesis constructs a prototype system on the basis of service reliability prediction theory and theory implementation under the background of big data,and provides an application demonstration of music service reliability prediction.The building of prototype system follows the steps such as requirement analysis,outline design and detailed design and implementation,and completes the function modules of user similarity threshold calculation,user group attribution,service recommendation and so on,which verifies the feasibility of the proposed algorithm and shows the combined theory of group-oriented service reliability prediction and the recommendation effect of its implementation method under the real application scenario in big data environment.