Research On Spatial Co-location Pattern Mining Based On Fuzzy Technology

With the continuous development of the global positioning system and the rapid popularization of mobile devices,the spatial data,which based on the location information,has growing explosively.Compared with traditional data,the semantic information of spatial data is more abundant,and also,the form and property of spatial data are more complex,which making spatial data mining more challenging than traditional transaction data mining.In order to discover useful knowledge from massive spatial data and guide human's production and life,spatial data mining emerged.Spatial co-location pattern mining is a branch of spatial data mining.Its main purpose is to mine the set of spatial features whose instances are frequently located together in space.Mining spatial co-location pattern plays an important role in spatial decision-making,intelligent city construction,environmental protection and other fields.The framework of traditional co-location pattern mining has many limitations.Firstly,the traditional co-location pattern mining framework uses a single proximity threshold to determine the spatial relationship,which results in the absence of the relationship between instances,and it also does not consider the influence of distance on the degree of relationship between instances.Secondly,the framework of traditional co-location pattern mining uses minimum participation to calculate the frequency of co-location patterns,which will lead to the absence of the co-location patterns with rare features,and the algorithm is very sensitive to minimum participation.In order to overcome the limitations of traditional co-location pattern mining,clustering algorithm can be used to cluster the features with high proximity.However,the index proposed by the traditional clustering method,which is used to describe the degree of correlation between features,is not consistent with the definition of co-location pattern.And the hierarchical clustering algorithm used in this method also has some problems,such as no redistribution,no overlap between clusters and so on.These problems will affect the accuracy of the clustering results,so the hierarchical clustering algorithm is not suitable for the mining of co-location patterns.To solve the above problems,considering that the proximity relationship is a fuzzy concept,this paper introduces the fuzzy theory into co-location pattern mining.Firstly,a new proximity measurement that can solve the traditional 0-1 proximity's problem is proposed.Secondly,the definition of fuzzy proximity is proposed,which has the properties of non-negative boundedness,symmetry and non-reflexivity,so that it can be well used in feature clustering algornthm.At the same time,a pattern mining algorithm based on fuzzy C medoids clustering algorithm,FCB,is proposed.Finally,extensive experiments on simulated data sets and real data sets prove the practicability and efficiency of the proposed mining algorithm.At the same time,it proves that the algornthm has low sensitivity to threshold and has high robustness.