Research And Application Of Spatial Clustering Method Based On Kulldorff Scan Statistics

Spatial scan statistics is a kind of space clustering method.The scan statistic method is different from the traditional spatial clustering method,which aims to detect the abnormal spatial distribution.The property features of space abnormal area is significantly different spatial objects.The scan statistic method detects whether there is a significance high risk likely cluster.The study means to detect and determine the location of the likely cluster,and the scope and degree of aggregation,which provide the basis and opinions for early warning and decision-making.The clustering method based on scan statistic has been widely used in geographic information science,geography,transportation planning,public health,ecology,sociology,economics,demography,criminology,disasters research and other related fields.Space and space-time scan statistic method was first proposed by Kulldorff.This paper researches the spatial,temporal and spatial clustering method based on Kulldorff scan statistics to detect the abnormal clusters,the main contents included below:(1)According to Tobler's first law of geography that the spatial interaction is universal,so we research a geospatial scan statistical method with interaction model,which can scan the space with the depth or breadth search mode based on interaction geospatial scan statistic to identify the significance abnormal spatial cluster.Then we test and evaluate the ability of the method by a variety of test data sets,and compared the ability of detecting abnormal spatial to Kulldorff circular,elliptical scan method.The method is also applied in detecting the abnormal distribution of social networks and track of the transport sector.(2)A space-time scan statistic method based on gravitational search algorithm was researched,which construct a cylindrical scan window and regard the space-time statistics as the optimization goal that can search the space-time domain according to the principle of gravity attractive,as result find the space-time abnormal cluster.The method is applied in detecting the time and space abnormal clusters with foot and mouth disease in Guangdong Province,and was compared with genetic algorithms,simulated annealing algorithm.(3)We research a Spatial Scan method based on Scale Space Theory,by introducing other attributes associated with the detected attribute to measure the spatial relationship between the properties under a single scale,whit spatial statistics unidirectional search,the highly relevant spatial units are merged in the process of in spatial scale changing.Finally we get the clusters at different spatial scales as result.Aiming spatial,temporal and spatial clustering anomaly detection based Kulldorff scan statistics the paper have research and application in the clustering method by considering the interaction in space,the optimization of time and space aggregation,the correlation of properties,and the related researches have certain theoretical and practical value.