Research On Cooperative Mechanism For Cloud Service Based On User Preference Community

The intensive economic benefits of cloud computing make cloud service become a new mode of software as a service. However, the uncertainty and complexity of quality of cloud service and the difference of users’ demands have brought many new challenges to the popularization of cloud service. How to select a service to be satisfied for the user and how to design a service according to the user’s preferences for the service provider are the two issues to be resolved that promote service trade. The goal of the thesis has two parts:one is to analyze the characteristics of the user’s preferences in the cloud and recommend a service to the user based on his preferences so as to improve his satisfaction; the other is to promote the service provider to improve the quality of service so as to suit the preferences of different user groups. Eventually a set of effective service cooperative mechanism is formed. The methods in the thesis will effectively make an optimum allocation of cloud resources and promote efficient trade between cloud users and service providers.In this thesis, a set of innovative approachs is presented to solve the problems of service optimization and service selection in the service trade by using community theory, game theory and genetic evolution theory. The related algorithms of each part are designed, and eventually the algorithms are deployed to the cloud platform, which implements a set of complete support system for service cooperative mechanism. The main work and innovations of the thesis are as follows:1. Put forward the concept of user preference community in the cloud and its discovery algorithmFirstly, the concept of user preference community is presented in the cloud by learning from the idea of community discovery. And then a clustering algorithm for community based on user preference similarity is designed by mining the evaluation information of cloud services. Finally an index is defined to evaluate the community structure based on users’ preferences, called community degree, so as to determine the optimal community structure. The experimental results show that the community discovery algorithm based on users’ preferences can effectively identify the user group with similar preferences in the cloud. So it can provide accurate decision-making basis to select credible users for service recommendation and design high-quality services based on users’ preferences.2. Propose a cloud service selection model based on community trustBased on stable user community structure, a service selection model based on community trust is presented to predict a service’s evaluation, which enhances the stability of recommendation users. The model includes two predictive methods:Intra-community predictive evaluation and inter-community predictive evaluation. The predictive decision parameter is used to automatically determine a suitable prediction method, to a certain extent, which overcomes the problem of data sparseness in service recommendation. The experimental results show that the community trust-driven service prediction has high accuracy, which is a feasible and effective method in application.3. Propose a game model between users and service providers based on QoSFirstly, a utility game model between service provider and user preference community is built based on QoS. And then the utility functions of users and service providers are defined and the conditions of game equilibrium between them are discussed. Finally an evolutionary equilibrium strategy based on QoS is presented with the goal of cooperation. The study of the model and strategy provides a basis of optimization for the strategy of service providers.4. Propose a QoS evolutionary method for cloud service based on user preference communityFor the preferences of user community, a QoS genetic evolution algorithm based on utility game model is designed to solve an optimal QoS strategy of the service, which can tradeoff the user’s satisfaction and the service’s revenue. The global search ability and efficiency of the algorithm are improved by the improvements of genetic operations. The experimental results show that the algorithm can effectively optimize the QoS from two aspects of users’ preferences and service cost, and helps to realize a win-win situation between service providers and users.