The Application Research On Spatial Data Mining In Surface Water Quality Evaluation And Prediction

Geographical Information System(GIS) is a major measure of spatial data managementand analysis;however, it can't automatically mine various potential spatial pattern, spatialrelation and spatial distributing discipline. Spatial Data Mining(SDM) is applied into spatialdata, and combined with the powerful GIS ability of spatial analysis and manipulation tomine connotative and potential spatial characteristic and spatial pattern from a vast ofnon-structured spatial data, finally serviced for scientific research and spatial decision.Spatial data and attributed data from water resource management are characteristic offar-ranging resources, various kinds, a large number of data, frequent change, so the need forpotential knowledge of applied research has been not satisfied by traditional GIS. SDM isused to manage and analyze these data, discovering useful knowledge and discipline,providing important dependence for water resource evaluation, program and decision.Taking Second Songhua River as study area, SQL Server-based spatial data warehouse isconstructed, and GIS, statistical data mining, visualized data mining, BP neural network areadopted to pretreat data, design water quality evaluation and forecast model, and analyze forspatial distributing discipline of water quality. Thereinto, the improved BP neural network isused to design the evaluating and forecasting model. In the network, 13 water evaluatingitems are selected as nodes in input layer;6 classes of evaluating results are selected as nodesof output layer with "0, 1" identified pattern. By continually practiced and compared,"13-9-5-6" double hidden layers with optimized training structure are confirmed. Later theprecision analysis proves that the model satisfies the need for water quality evaluation.This paper was divided into six chapters as follows: The first chapter dissertates theevolvement of SDM and water resources evaluation, expresses the main researching goals,technical scheme and so on. The second chapter introduces the situation of physicalenvironment, social economy and water resources exploitation and utilization in SecondSonghua River. The third chapter is to prepare data, included data integration, data clearing,data transformation and data reduction, then construction of data warehouse and knowledgebase of water resources evaluation. The forth chapter designs the improved BP neural networkwater evaluating and forecasting model, and validated its precision. The fifth chapter analyzesthe results based on visualized data mining, about spatial distribution discipline of SecondSonghua River water quality and it's affecting factors. In last chapter, this paper issummarized, the shortage of this model is pointed out and further works for it are expected.