Spatial Co-location Pattern Mining Based On Density Peaks Clustering And Fuzzy Theory

With the arrival of the era of big data and the popularity of mobile devices,mass of spatial data is flooded in people's lives.How to use these data to excavate interesting knowledge and improve the people's production and living conditions is a very important issue.Spatial data has the characteristics of high dimension and uncertainty,which makes the traditional data analysis technology have many limitations when dealing with massive spatial data.In recent years,with the gradual development of spatial data mining technology,processing and analyzing massive spatial data and applying them to practical applications has become an important goal of research.Co-location pattern mining is an important branch in the field of spatial data mining.The spatial co-location pattern is a set of spatial features whose instances are frequently located together in geographic space.The traditional co-location pattern mining method usually determines the proximity relationship based on the single proximity threshold given by the user.However,there are many limitations in actual mining process with the single threshold used.First of all,the“proximity”in real life is a relatively and fuzzy concept,using a single threshold is difficult to accurately describe the relationship between the all instances;secondly,because of the diversity of spatial distribution of the data,for the large density gap data,using a single distance threshold cannot obtain all proximity relationship,that affect the mining results;third,the instances of the proximity pattern will be repeated to calculated and checked,this process needs to consume a lot of time and storage space.In order to solve these above problems,this thesis firstly introduces the concept of traditional co-location pattern,clustering algorithm and fuzzy theory;second,this paper propose a new method for mining co-location pattern based on density peaks clustering and fuzzy theory,through introducing density peaks clustering and combining the fuzzy theory to achieve the overlapping partition of spatial data and proposes a series of definitions of fuzzy proximity relationship,fuzzy clique relationship,fuzzy participation index to establish a framework of mining co-location pattern based on density peaks clustering and fuzzy theory;third,this paper designs a DPC-MCP algorithm that can efficiently mining prevalent co-location pattern;fourth,the comparative experiments on real data sets shows the practicability and efficiently of the DPC-MCP on real data sets.