Design And Implementation Of Spatial Data Mining System (M-SDM) Based On MATLAB

Since the concept of spatial data mining was put forward in 1990s, the study of spatial data mining and knowledge discovery on massive data has gained the attention of people, and many research results have emerged, but they are mainly focusing on theoretical research and small-scale pilot study. The study on spatial data mining integration system facing to massive spatial data is weak, and on the other hand it is the urgent problem in practice. In view of this, this study discussed technical approach of spatial data mining from a new aspect, that is, designing and implementing a spatial data mining integrated information system based on MATLAB which has a strong numerical computation and visual graphic design ability, so as to provide technical support and solutions for spatial data mining from massive data.Using the designing concept of data analyzing software for imaging brain function——SPM (Statistical Parameters Mapping) for reference, this study combined MATLAB, GIS and SDM organically, constructed SDM system framework on MATLAB platform, integrated the major algorithms such as spatial association rule mining, spatial clustering analyzing and decision tree analyzing, and applied the system in land-use spatial database, aiming at enhancing the efficiency of massive data processing and enlarging the application of MATLAB in spatial data mining, spatial vector data processing and other aspects.This paper discussed research background of spatial data mining and status quo of current SDM systems, introduced related theories of spatial data mining in detail and the demand analysis of M-SDM, put forward the three-tier architecture for M-SDM design and development, realized the construction of system framework, designed spatial data mining Apriori algorithm, spatial fuzzy clustering and C4.5 algorithm at length, and developed the M-SDM system. The modules of this system are independent of each other, with a flexible and open structure. Moreover, specific examples were tested, and the feasibility and effectiveness of the system were verified, respectively in each module. As expected, valuable knowledge and rules have been acquired.The proposed technology and system architecture of spatial data mining system on the platform of MATLAB, the resolved key issues in system development and integration (such as database connection and use, SDM algorithm, visual evaluation of knowledge and rules, etc.), the developed spatial data mining integration system, as well as the successful application in land-use spatial data mining, have proved that it is a feasible research approach to construct a spatial data mining integrated system based on MATLAB. In addition, the research ideas, technical route, algorithm design and system integration can provide guidance and reference for other related research.