Study On Rainfall Erosion Prediction Model About The Loess Plateau Based On Data Mining

Rainfall is the main causation for soil erosion. To mitigate the jeopardy from rainfall erosion, our government has laid the crucial pivot of the nation's water and soil preservation research on constructing rainfall erosion prediction model, which is a important part of research on soil erosion, and it can establish the foundation for researching soil erosion in fixed quantity, as the main instrument to forecast the soil erosion quantum caused by rainfall. The factual measuring data about the Loess Plateau(the Yellow River drainage area) is introduced as analyzing element to endow the model with practical application value, and mathematical modeling method is used to figure out the equation expression of rainfall erosion prediction model.The rainfall erosion prediction model is a contributing part of soil erosion model. According to the theory of modeling the soil erosion, the research of this dissertation achieves analyzing basis using the method provided by gene analyzing model. Thus in this dissertation, the rainfall erosion prediction model is constructed in the light of rainfall erosion gene and the measuring data gained from soil erosion quantum.On the basis of understanding the mechanism of rainfall erosion and relative theory about soil erosion, this dissertation constructs the rainfall erosion prediction model with data analyzing as the instrument to achieve the aim of practical application. Concretely, the regression analysis is introduced as the main mathematical modeling method, combined with the data pretreatment technique of data mining. Thus the data is proceeded to accomplish effect of optimizing the model. SPSS13.0 is served as the main data analyzing software in the Windows XP application environment. The main achievements of this dissertation is as follows:(1) By analyzing the rainfall erosion data, combined with data pretreatment technique and actual meanings of the data, the data pretreatment algorithm is proposed for rainfall erosion prediction model. The analyzing results indicate that the model constructed with data proceeded by this algorithm behaves better than the one constructed with original data.(2) The multiplicative model for rainfall erosion prediction model is put forward by analyzing the rainfall erosion data with curve fitting technique and combining comprehension of Universal Soil Loss Equation (USLE).(3) The multiplicative model, with the method of linear transformation for it, is translated into expression form which can be figured out by linear regression analysis. The resolved result is transformed to achieve rainfall erosion prediction multiplicative model.(4) The model evaluation criterions used in this dissertation are all in the form of numerical value. So these criterions can be applied for quantificational evaluation of the results in comparing and evaluating the model. In addition, graphics are employed to analyze and compare results of the model. This method helps to find out problems in model solution.At last this dissertation summarized some questions and methods in the process of research and indicates the foreground of constructing and applying the rainfall erosion prediction model.