Spatial High Impact Co-location Pattern Mining

With the development of data mining research field,research work of data mining extensions to spatial database.With the support of satellite and remote sensing technology in spatial database,collecting and storing the abundant spatial data increasingly.The application demand for spatial data further promoting research and development of spatial data mining technology.In the geographic information system,geographic market,remote sensing,medical image processing,public health,national defense,ecological and environmental research,and other fields have been widely used.Different with traditional data,spatial data are usually related,namely the two space position of the object closer,the more likely having similar properties.The juxtaposition of space(co-location)pattern is a sub space feature set,their instances in space co-occur frequently.For example,Botanists discovered "semi humid evergreen broad-leaved forest"growing with 80%place having "blue" plants.In the study of mining spatial co-location pattern in the past,whether mined object is point object or extended object,without considering the influence of.spatial features and their instances,and also the evaluation index of interesting proposed such as PI(Participate Index)and CR(Coverage Ratio)do not reflect the differences of spatial features influence and their instances influence,spatial regions combination of instances.Therefore,this paper proposes high impact co-location pattern and mining algorithm.Firstly,we define a new concept the "effect",the definition of spatial instances effect,spatial features effect and the effect of the feature in the pattern are proposed.For evaluating high impact pattern how interesting rationally,we define the Effect ratio of the spatial feature in the pattern(ER:Effect Ratio)and the Effect index of the pattern(EI),comprehensively evaluating interesting degree of the pattern and the spatial feature in the high impact pattern.Secondly,the basic algorithm of co-location high impact pattern mining is proposed(HICPBA),and put forward the corresponding pruning strategy.Finally,through the experimental analysis and verification of the mining algorithm with pruning,also analyze the high impact co-location patterns mined are meaningful.