Research On Robustness Optimization Of Complex Networks And Its Application In Recommendation System

With the continuous development of science and technology,human's life has been greatly improved,which all depends on the normal operation of the infrastructure networks around us.How can we improve the reliability of these infrastructure networks? One of the tasks of complex network optimization is to solve this problem.This thesis,on the one hand,studies the optimization of network robustness,and optimizes the robustness and effectiveness of the network simultaneously by analyzing the topology of the network.On the other hand,this method of complex network analysis is applied to the recommender system,and thereby enhancing the reliability of the recommender system.The main contents of this thesis are as follows:1.An improved simulated annealing algorithm is proposed to improve the robustness and effectiveness of the scale-free networks.The existing network optimization algorithm either considers only optimizing the robustness of the network,or only optimizing the efficiency.In fact,these two aspects of the network are both important.For this reason,this thesis presents a new comprehensive evaluation function through combining robustness index and efficiency index together reasonably,and employs a novel evaluation model based on the rate of change to optimize the network structure iteratively,and ultimately achieves a robust and effective network.Besides,it is already pointed out that the topology of highly robust networks has a common characteristic that it presents a layered structure like an onion,that is to say,the nodes in the inner layer have a higher degree and then gradually reduced from the inside to the outside.Taking the onion-like structure as the heuristic information,this thesis constructs a heuristic random searching operator.Then,heuristic search and random search are combined to propose a hybrid search mechanism which relies mainly on heuristic search while taking random search as secondary reliance.Finally,based on the simulated annealing algorithm,an improved simulated annealing algorithm is proposed to improve the robustness and effectiveness of the scale-free networks by combining the proposed new objective function and the heuristic hybrid search operator.The experimental results show that the improved algorithm proposed in this thesis can greatly enhance the robustness and effectiveness of the initial network.2.The research on recommender system based on information core is a new direction occurring in recent years.The set of information core is composed of the users who can reflect the whole information of the recommender system.How to discover the information core users reasonably? On the basis of the existing Rank method,an improved method of finding information core is presented in this thesis through adding the information of similarity between users to the process of evaluating the importance of the users,and the method is carried out by combining the recommendation algorithm based on network propagation.The experimental results show that the improved method can discover a better information core.In addition,the further analysis of information core sets reveals that the average degree of the core users in it is not very large while the average degree of the objects selected by them is larger,and the diversity of the choice among these core users is also high.