Research On Spatial High-Utility Co-Location Core Pattern Mining Technology

In the era of big data,spatial data mining is used to help people extract potentially valuable knowledge from massive amounts of spatial data.Spatial co-location pattern mining is an important branch of spatial data mining.and it is based on the property that the more adjacent objects in the geographic location have the stronger relevance to mine frequently adjacent spatial feature subsets,while spatial high-utility co-location pattern mining is to combine the utility attributes of spatial features(non-spatial attributes)into the co-location pattern mining to more comprehensively support decision-making.Existing spatial high-utility co-location pattern mining follows the traditional spatial co-location pattern mining method of artificially setting a distance threshold when measuring the proximity relationship of spatial instances.This method is not scientific due to the randomness of the spatial instance distribution,and the algorithm designed accordingly has some problems such as unstable and unsuitable for unevenly distributed data sets.At the same time,the mining framework takes the utility values of all features into consideration in the pattern,which also is not applicable in some practical situations such as when commercial planning around the scenic spot,the utility of the scenic spot itself(such as tourism income)will be calculated into the pattern,the utility of the scenic spot must be far higher than the utility of the surrounding business pattern.So the high-utility pattern mined in this way is not necessarily reliable.In order to solve the above problems,firstly,the idea of k-nearest-neighbors is introduced into the spatial high-utility co-location pattern mining to measure the neighbor relationship between instances in a more objective and reasonable way in this paper,and the definitions of the k-nearest neighbor instance set of the spatial instance and the k-nearest neighbor feature set of spatial feature are given.Further,the concept of core element and core pattern are proposed.In order to measure the utility of the proposed core pattern,the paper formalized the definition of feature core participation instance set,the feature's core utility participation rate and the pattern utility index of the core pattern,which solved the problem that some feature utility should not be included in the calculation in spatial high-utility pattern mining.Next,a general framework of "Mining spatial high-utility co-location core pattern" is proposed and a corresponding basic mining algorithm is designed in this paper.The "grid" method and "sequence-tree" are used to generate candidate patterns.Due to the dissatisfaction with the anti-monotonic nature of the utility of the core pattern,global pruning cannot be performed,so in order to improve the efficiency of the basic mining algorithm,the paper proposed four pruning optimization strategies,which greatly improves the mining efficiency of the algorithm.Then,a large number of experiments have been performed on real data sets and simulated data sets to compare with the existing spatial high-utility co-location pattern mining algorithms PUI and UPI,which verify the practicability and superiority of the algorithm proposed in this paper.Finally,a prototype system for mining spatial high-utility co-location core pattern is developed in the paper.The system is simple to operate and has a concise interface.It can interact well with users and visualize the mining results in satellite maps.