Intelligent Prediction Model And Application Of Environmental Quality Based On Normalized Index Transformation

The current situation of water environment,air environment, ecological environment, hydrological and water resources environment which the mankind depends on for living are concerned with human survival and development, so is social civilization and harmony. Therefore, we need to protect, manage and planning the living circumstances of human being. then, there is an urgent need to establish a mathematical model for prediction of various types of environmental system. The purpose is to master environment quality present situation and its trend of development, provide data and methods to environmental quality assessment and prediction. So environmental prediction is the basis of environmental management department, but also the foundation of environmental planning.Currently, the prediction models which are widely used in different areas including autoregressive model, multiple regression analysis, stepwise regression analysis, uncentainty analysis based on fuzzy analysis, grey correlation analysis and set pair analysis, and other intelligent prediction models based on neural network,rojection pursuit and support vector machine.Those prediction model has its own characteristics, and obtained the certain results. A variety of statistical prediction models can reflect the stochastic features between the influencing factors and the forecasting variables, but its can not reflect the fuzzy, incompatible, uncertainty between influencing factors and the forecasting variables and only for large sample numbers modeling is meaningful; uncertainty analysis prediction model need to design different funtion for different factors, the function design and calculation will be complex when impact factors and the number of samples become large. intelligent prediction model has self-learning, adaptive and nonlinear mapping ability, so it can be used in high dimension, nonlinear predictive modeling, but the parameters of models which are need to be optimized are increasing with the number of factors.Therefore, when the impact factor is large, not only the structure and programmingmodel becomes more complex and increase the amount of computation and learning efficiency and accuracy of the models will be affected, so, the utility also subject to restrictions. This paper establish forward neural network model(NV-FNN),projection pursuit regression model(NV-PPR) and support vector machine model(NV-SVR)applied to prediction of different environment systems based on the the normalized indexes values that combine the thought of normalized transform with intelligent optimization algorithm. The combination of the normalized transform with intelligent optimization algorithm is a simplified model for the prediction of environmental quality, which opens up a new way for universal, standard and unified.This paper combines the National Natural Science Foundation of China(51179110,51209024) with technological infrastructure work of special projects(2011IM011000), established the intelligent prediction model of environmental system based on the normalized indexes values of influencing factors and the forecasting variables,The main innovated research results that have obtained in this paper are as follows:1) Proposed the design principles and methods of reference value and normalized transform formula for predictive variables and their influence factors of various environmental system(Chapter 2).2) Combined the thought of normalized transform with intelligent optimization algorithm, established forward neural network model(NV-FNN),projection pursuit regression model(NV-PPR) and support vector machine model(NV-SVR) based on the the normalized indexes values and analyzed reliability and accuracy of those three models.(Chapter 3)3) The three intelligent prediction models are applied to prediction analysis including water environment, air environment, hydrology and water resources environment which have different and variety factors and samples, then calculate the accuracy of each instance, and compared the prediction results with a variety of other prediction methods.(Chapter 4,5)4) The three intelligent prediction models are applied to prediction analysis including water environment and air environment which have the time series data,then calculate the accuracy of each instance, and compared the prediction results with a variety of other prediction methods.(Chapter 6)5) Compared the specification to transform the intelligent prediction model with a variety of other prediction model characteristics, conclusion the intelligent predictionmodel and prospected its application in related disciplines and fields.The main innovation points are as follows:(1) Proposed the design principles and methods of reference value and normalized transform formula for predictive variables and their influence factors of various environmental system, the reference values and specification transformation is provided with operation, specification and simplicity, so that on any number of factors predictive modeling can be simplified to only on an "equivalent" specification of factors predictive modeling, greatly reduced the number of modeling and forecasting the influence factor.(2) Combined the thought of normalized transform with intelligent optimization algorithm, established forward neural network model(NV-FNN),projection pursuit regression model(NV-PPR) and support vector machine model(NV-SVR) based on the the normalized indexes values,three intelligent forecasting model and make the prediction model of the structure to be simple, universal, standardized and unified and in a certain extent, improves the learning efficiency and accuracy.(3) Presented reliability analysis of three kinds of intelligent prediction models,and a new method of error correction is proposed to improve the accuracy of prediction model.