Research On Data Mining Application Based On Spatial Association Rules

With the rapid development of surveying and mapping science and technology,the methods about geospatial information collection,storage and management continue to improve.Geographic information data presents a large volume,multi-temporal and multi-structural characteristics.Geospatial data not only contains the basic attributes and spatial relationships of spatial objects,it also covers implicit information such as interdependence,mutual relations,restriction relations,symbiotic relations among spatial entities,and it is a natural data mining field.Deep mining of the hidden information of spatial data can improve data utilization rate and expand the breadth and depth of data utilization,which is of great significance to the development of various industries.Spatial association rule mining is currently a hot technology for acquiring hidden knowledge in geospatial data,and has excellent ability to acquire hidden information between spatial entities.In this paper,TOD(Transit Oriented Development,public transportation-oriented urban development model)site planning and location selection in Chengdu was took as the application.Based on the Chengdu Geographic National Conditions Monitoring Achievement Database,the spatial data mining technology is used to mine the association rules related to site selection,then the mining results are analyzed for rule integrity and feature combination characteristics and the TOD site-related knowledge model is constructed for TOD selection,which can make recommendations at the site selection.Meanwhile,based on the summary of the research practice,the design and implementation of the data mining module reduces the labor input of a large number of manual operations in the research and simplifies the difficulty of spatial data mining.The main contents of this article are as follows:(1)Based on the investigation on the application of TOD site selection in Chengdu and the demand for innovative application of geographic information by Chengdu Planning and Natural Resources Bureau,spatial association rule mining technology is used to mine hidden knowledge related to site selection.The composition of the original geographical elements within the coverage of the selected TOD site was counted,the correlation between the construction concept of the TOD site and each geographical element was analyzed,and then combined with the completeness of the existing basic data,seven spatial characteristics were Determined as input features for association rule mining.According to the rule generation principle of the association rule algorithm,the address clarity of 162 TOD stations planned to be planned in Chengdu and the integrity of the collected basic data at this point are sorted out,31 of which are selected as experimental sampling points Thirty-one unselected rail transit stations serve as control points,establish a basic spatial database,and establish a data base for final data mining.For the problem of low quality of the mining results caused by the spatial predicate conversion partitioning defects in the current spatial data association rule mining,the deficiencies of the traditional data partitioning methods such as isobaric and equal width were discussed and summarized.By comparing the advantages and disadvantages of various discrete partitioning methods,combined with the characteristics of the actual collected data,the K-means clustering method was finally used to partition the basic data,and the partition comparison table was established for the selected 7 spatial features,which overcomes the problem of low data similarity in the quantized data partition family.(2)Use data mining software to mine the transaction database to obtain interesting knowledge related to TOD site selection.Aiming at solving the problem of effectively filtering uninteresting rules,through the design of a combined confidence and support experiment,20% confidence and 20% support were determined as the optimal input parameter values for the current mining experiment,which effectively avoids the uninteresting rules in the mining results and also ensures that the interesting rules are not lost.Using WEKA data mining software to mine the spatial transaction database,a total of 59 spatial association rules related to TOD site selection were obtained.Through the overall interpretation and combination analysis of the obtained rules,a knowledge model of site selection was established,which can provide certain reference suggestions for the site planning of TOD sites in Chengdu.(3)Through the practice of spatial data mining research,the defects of data loss,confusion,and repeatability caused by the conversion of the mining process during the research process were recorded and organized.The disadvantages caused by a large number of manual operations in the data mining process were analyzed.The result showed that it is necessary to develop the data mining module.Combining the functions of each sub-process and the standards of data input and output in the process of data mining,the data mining module is divided into 4 sub-modules for data collection,data discretization,data mining,and achievement display,and finally the development of the data mining module is designed and completed.To a certain extent,the data mining module reduces the labor cost input and operation error probability in spatial data mining research.