The Implement Of An Analysis System For Networked Seismic Big Data And Its Applications On Earthquake Predictions

Earthquake network is a crucial approach based on complex network to study seismic system,where nodes represent the cell of geographical region containing earthquakes and edges represent the correlation between earthquakes considering their important information about occurrence place,time,magnitude,etc.Earthquake network is proved to be scale free and small world.Scientists can study the structure and function of the system as a whole,explore its internal connection,and find the essential features of earthquakes.Many present complex network analysis tools are not suitable for the study of earthquake network.Based on this point,we design an analysis system for networked seismic big data in this thesis.First,the basic theoretical knowledge of the system has been introduced.Then,the requirement analysis and overall design of the system has been done.Specially,we analyze and solve the problems occurring at the probabilistic inference module.This module uses Bayesian network to represent the dependency correlations between nodes.The large scale of the network and the closely connection between nodes make huge computational complexity when we compute the conditional probability.At last,we make a prediction of earthquake events and use the prediction precision to prove the validity of the method.Next,the system implementation has been introduced.The system is divided into 5 modules,and mainly achieves the following functions:It produces corresponding network topology according to the different data source,the size of the cell and the network construction method;it computes 10 characteristics of the network;it achieves the function of evolution analysis of different parameters;it achieves the function of correlation analysis between different parameters;it achieves the function of probabilistic inference and event prediction according to the correlation between nodes.At last,we take a functional test on the basis of the system.We use the system to complete the study of topology evolution and probabilistic inference.In the test of evolution,we study the evolution of topological characteristics at the occurrence of great shocks.The result is that the scale,order degree and community structure represent some differences before and after the great shocks.We analyze the earthquake network constructed from the data in California,1992,using multiple parameters at the same time to better understand the changes of the networks.In the test of probabilistic inference,it is found by a large number of experiments that the prediction precision is high when we choose the year contains great shocks as training set.Furthermore,the prediction precision can be improved when the time range of training set is increased,especially when the test set contains great shocks.Moreover,we use the earthquake data of California,2015,to apply the prediction.The system designed and realized by us in this thesis has already been running in our laboratory.The actual running result shows that our system is undoubtedly of considerable realistic significance and has application prospect for the study of earthquake network and analysis of earthquake events.