Research On Technologies Of Cloud Service Selection And Recommendation Based On QoS

As a new type of Internet service model,cloud computing can provide shared software,hardware,and information in the form of services to users and other equipments on demand,so as to achieve flexible construction and high integration of military information resources during the integrative joint operation.With the further development of cloud computing,the number of cloud services on the Internet has experienced explosive growth.How to select the service that meets the user's needs and recommend the service in line with the user's preferences is a research hotspot in the current cloud computing field.On the basis of in-depth analysis of cloud service features,this paper focuses on the issues of dynamic multi-binding,accuracy,and diversity existing in the current service selection and recommendation technologies,so as to fully improve the user satisfication.The main contents and contributions are as follows:1.In order to solve the uncertainty of QoS value for cloud services and unreasonable computation of QoS attribute weight,a service selection method based on dynamic QoS and subjective and objective weighting is proposed.An interval QoS model is established to present the uncertainty of QoS values and ambiguity of user QoS requirements.On this basis,TOPSIS(Technique for Order Preference by Similarity to Ideal Solution)method for MADM(Multiple Attribute Decision Making)problems is used to compute the objective weight of QoS attribute.By combining with the user subjective preference,the comprehensive weight of QoS attribute is finally computed.By taking into account of both objective QoS data and user preference,the method effectively improves the accuracy of service selection.2.To solve the problem that the composite service selection is not efficient and the functional constraints among services are ignored in the current cloud environment,a multi-constraint service selection method based on decomposition of global QoS is proposed.The quality level and learning mechanism in the belief space are introduced to improve the cultural genetic algorithm,so as to decompose the global QoS constraint optimally.By combining functional constraints with QoS constraints,candidate services are filtered and the optimization scale is reduced greatly,enabling the algorithm to quickly select a global composite service,which satisfies the real-time requirements of users.3.For the problem that the trust relationship between users is sparse and recommended services do not fully meet user preferences,a trust expansion and listwise learning-to-rank based service recommendation method is proposed.By introducing the sorting information of services,the probability distribution model is used to compute the user similarity,which is more accurate than traditional algorithms.Then a trust expansion model is proposed to fully mine the trust relationship between users,and by combining with user similarity,a trusted neighbor set is constructed to resist malicious ratings.An optimal ranking model is finally trained by listwise learning-to-rank algorithm,which can accurately recommend services that meet user preferences.4.Concerning the sparsity of user preference information and lack of diversity in recommended services,a service recommendation method based on random walk and diversified graph ranking is proposed.In order to alleviate the data sparsity,random walk is presented in the user network to find more similar users.By constructing the service graph model,the service recommendation problem is transformed into finding the optimal node set.The diversified ranking algorithm in graph theory is then introduced to balance the accuracy and diversity of recommendation results,and achieves personalized service recommendation.