摘 要：隨著成千上萬種化學物質的產生并排放到水中，化學物質的環境危險評價越來越重要。雖然部分有機毒性物的實驗測定現在己能做到，但由于這些毒性物的數量非常眾多并且存在其他一些潛在的有機物仍未進行實驗測定，所以定量預測環境中有機毒性物的毒性仍是十分重要和必須的。本論文主要討論徑向基人工神經網絡（Radial Basis Function Network，RBF）用于胺類有機物毒性識別問題，內容包括有機物的環境影響評價的意義，胺類有機物對環境的影響分析，RBF人工神經網絡建模的思想、結構和算法，以及胺類有機物QSAR（定量結構-活性相關，Quantitative Structure-Activity Relationship）分類問題RBF實現的基本設定。實驗部分是以胺類有機物中的各個變量為基礎做出建模，并使用MATLAB軟件進行相關程序的運行，進而做出數據處理和結果分析，計算其結果的準確性，并和多元線形回歸方法進行比較。
關鍵詞：環境影響評價；胺類有機物；毒性識別；人工神經網絡；RBF
Abstract：With thousands of chemicals producing and releasing into the water, chemical eva luation of the environment is more important. However, some of the experimental determination of organic toxic now can be done, but because these were the toxicity of a very large number as well as the existence of other potential organic matters that have not been measured yet, therefore, the quantitative estimation of the toxicity of organic toxic material is still important and necessary. This report mainly discussed RBF artificial neural network (radial basis function network. RBF) for the identification of toxic organic amines, including environmental impact assessment of the significance of the organic matters, analysis of the environmental impact of organic amines, RBF artificial neural network modeling ideas, structure and algorithm, and organic amines QSAR (Quantitative Structure-Activity Relationship) RBF classification, the basic set. Experimental part is based on various variables of the organic amines and make modeling, use the software of MATLAB for the operating the related procedure, then make the data processing and result analysis, calculate the accuracy of the result, and compare with the way of multiple linear regression.
Key words：Environmental impact assessment; Organic amines; Toxicity identification; Artificial neural network; RBF;
