Spatial Data Mining In Valley Based On Genetic Algorithm

China is insufficient in freshwater resources. It is significant to river resources rational utilization and exploitation, which can fit the whole national economic development. In order to realize sustainable development of freshwater resources, we must grasp the inner rules and inherent relation of each essential character, so that watershed data from all ways is deeply analyzed, and characteristic-information which reflects incidence relations between hydrology and regime on watershed space is obtained. However, the researches relied on spatial data are few and also a challenging problem. The technology of Spatial Data Mining provides the possibility to solve the problem. However, spatial data mining in basin encounter many difficulties for complexities of watershed's spatial data and traditional hydrological model.Based upon the national natural science foundation, a model of space-series spatial data mining in basin is proposed in this thesis, which is based on the principles and methods of SDE and overcomes the deficiencies that traditional hydrological model was separate from space characteristics. More mature Genetic Algorithm is adopted to build the model which based on space characteristics, and try to find the inherent relation of hydrologic data from observation stations. The case of Qingjiang River Basin proves the feasibility of this model in transect reduction and hydrologic prediction. Finally, combined with the technology of Geographic Information System, the results of space characteristics data are spread in the basin GIS, which provides more visual information for scientific management and decision. Main research work is listed as follows:The procedures and flow of spatial data mining in basin are set up based on the general process of SDE, combined with the particularity of watershed's spatial data.Considering the complexities of watershed's spatial data, a model of space-series hydrological prediction model is proposed instead of traditional hydrological model based on time-series. The model uses least parameters and variables to reflect the calculation and multiply problems in hydrologic data, and avoids the complicated relations among the influencing factors of spatial information, which is important for building the spatial data mining model in basin.The processes of spatial data mining are discussed in detail. Genetic Algorithm, a non-numerical optimization algorithm, is used for the model. The ability of global searching and solving non-linear objective guarantees the feasibility and effectiveness of the model.