Research On Evolving Model And Network Characteristics Of Adaptive Hypernetworks

A large number of complex systems in reality have a characteristic that there arerelations among multiple nodes simultaneously. But general complex networks cannot describe this character well. However, the hypernetwork based on hypergraphstructure has the characteristic that allows a hyperedge to connectmultiple homogeneous or inhomogeneous nodes, so that it can better describe the characteristic of complex network. Research on evolving model of hypernetwork will promote the understanding of the topological characteristics and evolving regulations of complex systems with a structure of multiple nodes. Combined with the NaturalScience Fund Project Research on social-driven context-aware personalizedinformation service in ubiquitous computing environment (project NO.:71471165)and the Social Science Planning Project of Ministry of Education of ChinaPersonalized information service research in mobile environment a context-awareand ontology based approach (project NO.:14JYC870010), this paper studies theevolving model of hypernetworks, and the main work of this thesis is as follows:(1) Two typical networks, scientific collaboration network and social taggingnetwork, are selected as empirical data. The statistical characteristics of the twonetworks is analyzed in this paper, including the growing trend of the networks withthe time, the node hyper-degree distribution, the relationship between the nodehyper-degree and the time of node first-occurrence. The result shows that thehypernetworks have a power-law node hyper-degree distribution, and the power-lawexponent is not only r≥2, but r<2. Also, the corresponding relation between the nodehyper-degree and the time of node first-occurrence doesn t exist, which shows afitter get richer phenomenon in real networks.(2) According to the statistical characteristics of real networks, preferentialattachment mechanism of combining the fitness with hyper-degree is proposed, andan evolving model of adaptive hypernetwork is constructed and simulated underdifferent fitness distribution function. Drawn from the results of the simulation, the feature of power-law distribution is more obvious when the fitness follows thepower-law distribution, which provides the basis for selecting it as the fitnessdistribution function of the model.(3) The power-law distribution is determined by the fitness of nodes in realnetworks. Also, the statistical measurements, such as the node hyper-degreedistribution, the clustering coefficient and the average path length, are analyzed forsimulated networks and real networks. The empirical results show that simulatednetworks and real networks have high consistency of the statistical characteristics,which proves the model does better in describing the evolution mechanism of realnetworks and explaining the fitter get richer. Also, this paper proposes and validatesthat the parameter of fitness distribution function meets the relationship=(L/M)+1, where L/M suggests the node growth rate of an evolving network.