Research On Spatial Data Mining Techniques Based On Mineral Belts Foundation Database

With the fast development of spatial data obtained technology, spatial data increase up rapidly.There are many connotations of knowledge and rules need to be excavated in the spatial datadatabases. So it resulted in the birth of Spatial Data Mining (SDM), a technique that scans andfinds desired knowledge from spatial databases.Spatial database is an aggregate of spatial data and attributive data, which describes spatialobjects, and also the database management system of collecting, storing, managing, searching, analyzing, expressing spatial data and attributive data.SDM, also known as Spatial Knowledge Discovery from database, is a method of distillinginterested spatial patterns and characters, universal relations of spatial data and no-spatial data, aswell as other connotative characters of datasets from spatial databases. It is an extension of datamining in databases, which promotes GIS to become more intelligent the appliance of spatial andintegrated.Mineral resource forecasting is the application core of the SDM techniques using in mineralresources fields. Multiple geo-data information is different from other data, because it has spatialattributes, in addition to these data has obvious implications, there is an implied rich meaningwhich shown beening analyzed and mined. At the same time, increasingly rich data to a certainextent over the earth scientists" capacity of being able to handle. So, from these massive datafound knowledge to make the needs of data analysis and SDM will be combined. The foundationdatabase for resource assessment about importance mineral potential belts of China centralizedmanages geo-data of 14 important mineral belts, including geography data, geology data, mineraldeposits data, aeromagnetics data, gravity data, geochemistry data and remote sensing data, withthe unified data model and data standard. For making full use of those data, we will study how toautomatically mine information from massive data and reveal spatial correlativity and hidegeo-rule of associated geo-data which is characteristically multi-source, multi-type, multi-factorand multi-quality, especially, each subject of geo-data has different integrity, data format andspatial scale. Then, we can give better support for prediction and assessment of mineral resourceand make full use of foundation spatial database.In the evaluation of mineral resource, it is difficult to use a mathematical model to expressprediction model, because the requested data has many types, great otherness and the ore depositis characterized by complex conditions. So, the use of artificial neural networks can better solve the above problem. In the geo-information fields, ANN will be widely used for it has strongability to learn, with the functional characteristics of correcting the noise and distortion signal.This paper discusses the feasibility and superiority of the BP model in the evaluation of mineralresource. Firstly, BP network arithmetic was programmed and the BP model of mineral resourcesassessment was established. Then, using selected sample to train BP, and the geological variableweights will be recorded when the BP network becomes stable and convergent. At last, user canevaluate the forecast cells of the study area and delineate the potential mineral areas using thetrained BP networks. Then, the aim of implementation of the study area's mineral resourcesevaluation and prediction can be achieved successfully.Finally, the spatial data mining system named MPBC_DataMiner has been designed andprogrammed. MPBC_DataMiner basing on the better basic spatial database of mineral potentialbelts of China is designed guided by mineral prediction theory, combining the latest componenttechnology, spatial database and spatial data mining techniques.This system is seamlesslyintegrated with MPBC (Database Management System for Mineral Potential Belt of China)which is the database management system for basic spatial database of mineral potential belts ofChina. We choose VB, ArcObjects, geodatabase data model which is the third generationgeographic spatial data model and spatial data engine ArcSDE (For SQL Server) to program thedata mining system. The MPBC_DataMiner system focuses on achieving a fractal algorithm andBP neural network algorithm, which provides users with a data manipulation and analysisplatform.In the end of this paper, the geo-data of one study belt was chosen for certificating thefeasibility of the system. Preliminary results show the system implements the functions ofaccessing multi-source geo-data from massive spatial database, conversion, mining, and dataanalysis and visualization, which is also able to integrate multi-source spatial data and mineknowledge, make good mineralization prediction.