Research On Methods Of Spatio-Temporal Association Analysis Of Social Security Events Based On Global Terrorism Database

Under the current national and international circumstances, improving social security event response capacity has become an urgent need of national governments. Analyze the association relationship of social security events by using the method of spatio-temporal association analysis could effectively improve the ability to cope with emergency events. Spatio-temporal association rules mining, spatio-temporal correlation analysis and visual analysis are main analysis methods of spatio-temporal data association analysis. Apply the method of spatio-temporal data association analysis to the research field of social security events combined with the property and structure characteristics of social security events’data, analyze potential relationship of social security events and forecast the overall development trend, then prevention and rapid response capability for social security events can be effectively improved, and then provide a reliable basis for manpower and resources department and making response measures. Using the data provided by Global Terrorism Database, this dissertation explores association analysis methods for social security events from three aspects as association rule mining, spatio-temporal correlation analysis and visualization analysis, and use experiments to verify the reliability of the proposed methods, the main work and innovations are as follows:1. Analyze the time, space and property characteristics of social security events, an event analysis and forecast model has been built and suitable data table structure for association analysis has been designed.2. In aspect of spatio-temporal association rule mining, research has been divided into two parts, non-time sequence association rule mining and time sequence association rule mining. In research of non-time sequence association rule mining, based on the FP-Growth algorithm, put forward an association rule mining method FP-Growth-T with time stamp, this method does not require pre-specified timing constrains and can earn results with time characteristic in the whole time region. The effectiveness of the method has been confirmed with real data from Global Terrorism Database. In research of time sequence association rule mining, introduce temporal autocorrelation function into spatio-temporal association rule mining to solve the problem of vagueness and subjectivity of time predicates. The reliability of association rule results can be greatly improved using this method which confirmed in experiments.3. In aspect of spatio-temporal correlation analysis, the spatial distribution and accumulation characteristics of social security events have been analyzed by using spatial autocorrelation function, the scope and propagation mechanisms of social security event in time domain have been analyzed by using time autocorrelation function, and the correctness of analysis results have been verified in methods of statistical analysis. According to correlation analysis, the factors social security events suffer from have been listed, and an event prediction model based on Bayesian network has been put forward in order to analysis and forecast the overall trend of events.4. Based on the visual characteristics of social security event, the scope and expression characteristics of variables and visual representation methods have been explored. Research on visualization of spatial and temporal positioning, and use statistical charts and statistical map to show the spatio-temporal distribution, property comparison and development trend of social security events as well as guiding readers to analyze the potential characteristic, the interrelationship and the overall development trend.