The Research On Case Spatiotemporal Data Mining

China is in a stage of social transformation where various social conflicts intensified and public safety situation is grim. Therefore strengthening emergency management and enhancing decision-making capacity of rapid response to urgent public emergency become a major strategic task, the basic content of which includes introducing modern data analysis techniques, extracting and mining spatiotemporal information of public emergencies from the thematic data warehouse of comprehensive emergency management, providing decision-making basis for emergency management departments and the public. Take case data of Fuzhou from 2005 to 2007 as sample, integrating data analysis techniques such as spatiotemporal correlation statistic, spatiotemporal clustering and spatiotemporal prediction, this thesis carried out the research on case spatiotemporal data mining. The main content and results are as follows.(1) Based on the summarization of development and research status of spatiotemporal data mining, taking public safety and comprehensive emergency business as demand-led, we combine case spatiotemporal data characteristics, spatiotemporal data models and existing researches in the field of spatiotemporal data mining, explore the methods and processes of case spatiotemporal information extraction.(2) As for the case spatiotemporal point data organized in the spatiotemporal cube model, we design a spatiotemporal hierarchical clustering algorithm based on Delaunay Tetrahedron Network and propose a cluster validity index for it, conduct the empirical research on case spatiotemporal clustering analysis on the basis of clustering trend information according to spatiotemporal correlation analysis, successfully evaluate the clustering results quality with the cluster validity index. Experimental results show that the algorithm can effectively process the case spatiotemporal point data, has low time complexity while been able to discover the correct clustering results, and can achieve spatiotemporal anomaly detection.(3) As for the case spatiotemporal polygon data organized in the sequence snapshot model, we introduce the commonly used predictive model STARMA from the spatiotemporal sequence analysis to study case spatiotemporal trends, conduct the empirical study on case spatiotemporal prediction analysis based on the temporal stationarity of case spatiotemporal sequences according to spatiotemporal correlation analysis, and comprehensively assess the fits and predictions accuracy of the model with five kinds of statistical indicators of analysis results. Experimental results show that the model can effectively process the case spatiotemporal polygon data with temporal stationarity, and can achieve spatiotemporal prediction and trend discovery.