Study On The Prediction And Assessment Methods Of Water Environment Quality Based On Support Vector Machines Theory

Innovative work is done on modeling and algorithm of Support Vector Machine (SVM) for the water environment quality prediction and quality evaluation in this dissertation. SVM plays a leading role in the sciences for complex non-linear phenomena and artificial intelligence. Researches on its application in the planning of the water environment quality prediction and quality evaluation are still in the preliminary stage in the world. On the basis of a comprehensive evaluation and analysis of the present situation of the researches in the water environment quality prediction and quality evaluation, and on the basis of a careful exposition of the basic principles, the algorithm and the varied pattern features of SVM, this dissertation gives an application of SVM approaches in the water environment quality prediction and quality evaluation, which, as the first attempt of its kind, can help to achieve a higher level in the application of artificial intelligence in this field.In fact, the water environment quality prediction and quality evaluation is subject of classification and regression. Traditional statistics is based on assumption that samples are infinite, so are most of current machines learning methods. However, in many practical cases, samples are limited. Most of existing methods based on traditional statistical theory may not work well for the situation of limited samples. Statistical Learning Theory (SLT) is a new statistical theory framework established from finite samples. SLT provides a powerful theory fundament to solve machine learning problems with small samples.Support Vector Machine (SVM) is a novel powerful machine learning method developed in the framework of SLT. SVM solves practical problems such as small samples, nonlinearity, over learning, high dimension and local minima, which exit in most of learning methods, and has high generalization. Currently, being the optimal learning theory for small samples, SLT and SVM is attracting more and more researcher and becoming a new active area in the field of artificial intelligent and machine learning.As a new subject, the research for SVMs only starts recently abroad. People didn't notice it until recent year in our country. Although SVM has a good system of theories, there are still many questions to be answered about its applications. Also there is much space to perfect and develop its theories.The main contributions are following:1) Introduce the background, development and the current research in water quality supervision and evaluation methods for water environment. Discuss the feasibility and innovation of the application of SVM in the water quality prediction and evaluation research from the angle of objectivity and fairness.2) Elaborate the essential problems and main contents of statistical learning theory. To explain how to apply the statistical learning theory in practical problems, firstly introduce the basic concepts and basic theories, and then focus on the basic theories and methods about linear and nonlinear standard support vector and also the standard support vector regression machine.3) Discuss two support vector machine algorithms and their variants in classification. The analysis and study on these algorithms shows the advantages, disadvantages and applicable range of each algorithm, which makes theoretical preparation for using SVM classification theory in water environment quality evaluation.4) Focus on the introduction of the fundamental principles of SVR for curve fitting, and then analyze several SVR models, includingε-SVM, v-SVM, LS-SVM, W-SVM and SVM based on linear program. After that establish the SVM model constrained by single parameter ofε?non-sensitive function, which verifies the equivalence of single parameter regression model and standard regression model. Some numerical experiments are conducted. Finally discuss the criteria to estimate the goodness of SVM algorithm, which perfects the optimum theory and decision-making of SVR.5) Summarize the existing multi-class classification methods, including one to many method, one to one method, hypersphere multi-classification method etc. Compare their advantages, disadvantages and performance and prove by experiment. A fuzzy iteration SVM multi-value classification algorithm based on decision binary tree is given combined with pre-draw SVM algorithm. The proposed method overcomes the problems of other existing methods, enhance the training and decision-making speed. Also it solves the problem of undivisible region for SVM multi-class classification method with satisfying results.6) In this paper, a new SVR algorithm is put forward to solve the prediction problem of contamination concentration regarding the supervision of water quality parameter average concentration as a time series prediction question. The supervision data of pollution material concentration is performed regression estimate analysis using SVR. Compared with the result of BP network method, the prediction model established by SVR method can make full use of the distribution features and the prediction result fits the practical condition better. Moreover, SVR is better than BP network in integrating performance with great learning ability and generalization capacity.7) According to the principles and requirement of water evaluation, a new method of iteration SVM multi-value classification algorithm based on decision binary tree is presented. Several simulations demonstrate that compared with the existing methods, the number of SVMs need to be trained is less by using the new method, the speed of training and decision is fast and the region that can not be classified does not exist again. It is used to set up a water quality evaluation model for Beijing lake reservoir. The analysis result shows that with numerous assessment indexes, SVM model can quickly produce assessment results with a high accuracy, and, as it can work with both qualitative and quantitative indexes, has wider areas of application.The innovated points are mainly listed as follows:1. This dissertation gives an application of SVM approaches in the water environment quality prediction and quality evaluation. As the first attempt of its kind, can help to achieve a higher level in the application of artificial intelligence in this field.2. Discuss the criteria to estimate the goodness of SVM algorithm, which perfects the optimum theory and decision-making of SVR.3. Extend the applicable range of SVM, a new and effect method of iteration SVM multi-value classification algorithm based on decision binary tree is presented to the water environment quality evaluation (CH 6).4. A new effective method is put forward to evaluate the water environment quality prediction based on the study of SVR algorithm (CH 7).It is a new trial to apply SVM theory and algorithm in the water quality prediction and evaluation, and achive some satisfying results. This method is small sample needed, simple in computation, fast, accurate, practical and and coincidence with the objective features of the research objects interest. It can also find the optimum compromise between the model complexity and learning ability according to the limited sample information such that it has high generalization. This paper will extend the applicable range of SVM.SVM technique will become mature as the development in the research method and hardware condition. SVM will be a very important research topic in every research field due to its special characters and will stimulate great progress in scientific technique.