Dynamic Simulation And Prediction Of Urban Sprawl Based On Bp Neural Network

Urban sprawl is a dynamic, non-linear, more complex feedback loop system, the traditional linear, non-linear system identification in the theoretical study and practical applications has a lot of limitations. BP neural network has the function of approach of any non-linear mapping through learning, BP neural network applying to non-linear systems modeling and forecasting has obvious advantages. This dissertation based on the urban sprawl and BP neural network of the home and the abroad, choosing Tai'an City as the regional study, Study the dynamic change process and the laws of the land use in the urban sprawl process and obtained the following results.Analysis and superposite all of the Tai'an City of urban land use map using ArcGIS software , revealing the law of urban sprawl from structural changes in land use, land-use changes in the type, built-up area in different periods of land-use change there aspects. Analysis the natural geographic, socio-economic, demographic development, transport and land policy in the five-driving factor on the impact of urban sprawl, determine the impact of urban sprawl of the dominant factor. Combinate ArcGIS software stack analysis and functional analysis of the buffer zone, extract the five variables distance, four neighborhood variables from the geographical and transport factor, and for the first time socio-economic, demographic development and land policy would be vectored, output the raster data formats, to participate the dynamic simulation of urban sprawl process, improving the accuracy of the simulation, clarifying the causal link between urban sprawl and the various driving factors.By analyzing and using the limitations of MATLAB toolbox in the neural network to establish urban sprawl forecast model and of the advantages of VB programming language, MATLAB will be made in the neural network function and Visual Basic 6.0 for the establishment of integrated urban sprawl forecast model approach, VB has the visual interface and MATLAB powerful data analysis capabilities, the combination of urban land-use planning and macro-control has provided a new workable theories and methods. The use of the establishment of urban sprawl model, simulating Tai'an City's dynamic process of urban sprawl, Forecast the urban land use situation in the 2010. Given the built-up areas superimposed map from 1980 to 2010, The process of expansion and law of Tai'an City was revealed well.The proposal stems from Supported By Open Research Fund Program of the Key Laboratory of Geomatics and Digital Technology, Shandong Province " Theory and method of the urban land use changing simulation on neural network" (SD060805) and the National Natural Science Foundation project "Spatial Data Mining Mine Disaster The neural network theory and method "(40671149).