Research On Construction,Analysis And Node Relation Prediction Methods Of Earthquake Data Relation Networks

Based on immense computing power of model computers,the research of complex network fuses topology,graph theory,nonlinear science,modern statistical physics and other fields.The methods of this discipline consist its own hierarchical structure.Whether protein structure of biological network,or the boundless cosmos,they both can be considered by the same networked model,which contains nodes and edges.The analysis of topology structure and dynamics process for these networks greatly improves the understanding of complexity of the real world.Therefore,complex network is becoming the focus of present research.Earthquake disaster frequently occurred in whole world at the 21st century.On the one hand,based on the historical research results,scientists speculated and proved the nonrandomness and nonindependence of seismic events.On the other hand,with the great increase of detectable seismic data,it is possible for the scientists of other fields to study seismology.Based on the complexity of seismic system,they suppose the seismic event as individual and complicated relation as the relation between individuals.They study the networking method,topology characteristics,and dynamical behavior of networks by following the general process of complex network research.Based on the space-time complexity and the cause and effect relation,this thesis studies the seismic data networking method,analyzes the topology characteristics,analyzed the network from the perspective of space and time,infers and predicts the nodes relations of earthquake network.Under the hypotheses of space-time cause and effect relation,we try to deeply understand the law of seismic system and provide a new approach for earthquake prediction.The specific topics of this thesis are as follows:Firstly,this thesis analyzes the lack of space-time cause and effect relation on the classic earthquake data networking approach.Based on the space-time complexity,it supposes the space-time influence domain and provides an earthquake network construction method.By using the earthquake data of south California,Greece and Japan,starting from the basic characteristics,this thesis analyzes the power-law distribution of degree,small-world and community structure.The results shows that earthquake network is heterogeneity,highly-clustered and independence.Secondly,to understand the variation of node position in topology structure,this thesis first combines the node centrality method with the space analysis of earthquake network.It studies the degree centrality,betweenness centrality and k-core of earthquake network.By k-core decomposition,it proves the hierarchical characteristic and found the connection feature among different shells and heterogeneity of nodes corness.It proves the highly-cluster feature of highest-core nodes and explains the physical implication by the space distribution of great shocks.Thirdly,based on observation of earthquake network evolution,unusual evolution of earthquake network topology is detected and defined.This thesis studies the long-term and short-term unusual evolution.By analyzing the long-term evlution,it studies network scale,average degree,average path length,clustering coe:fficient,network entropy and coreness and explains the physical implications from different parameters.By analyzing the short-term evlution,it studies 3 great shocks of south California combines the parameter of energy.The results show that the emerging nodes and the coreness changes of reserving nodes is the reason of unusual evolution.Finally,this thesis provides an earthquake node relation inference and prediction method based on Bayesian network.Due to particularity of the seismic data and the principle of Bayesian network,it extracts the simplied network and gives training and testing method of node relations.According to accuracy and recall rate,the results show that this method is acccurate in specified time and space domain and found that the accuracy has certain connection with great shocks,time spans of training sets and magnitude limitation of testing sets.The given methods broaden the mind of seismic research.Our research proves that they are necessary for theory and application.