运用回归分析与人工神经网络预测含硫芳香族化合物好氧-环境科学.pdfVIP

1998 1 ENVIRONM ENTA L SCIENCE Jan. , 1998 * 张爱茜　韩朔睽　沈　洲　马　静 ( , 210093) , , . , , , 0. 00102, 0. 01591. , , , , . Prediction of Aerobic Biodegradation Rate Constants - of Sulfur Containing Aromatic Compounds Using Regression and Neural Network Techniques Zhan A iqianHan ShuokuiShen ZhouM a Jin ( Department of E nvironmental Science and E n ineerin ,Nanjin U niversity ,Nanjin 210093) AbstractIn order to evaluate the ability of artificial neural netw orks to predict the biode rad- ability of or anic compounds, the codes of function roups w ere employed as structural descrip- tors to fit and predict the pseudo-fir st-order aerobic biode radation rate constants of a batch of - sulfur containin aromatic compounds usin multiple linear re ression and error back propa a- tion neural netw ork models, respectively . T he results indicated that the neur al network has hi her ability to solve the complex problems affected by confoundin structure factors such as biode radation process due to its ability to conclude inter actions betw een roups automatically . - 0. 00102, The mean square err or of the netw ork is w hich is much low er than that of the linear model 0. 01591. Ke wordsartificial neural netw or ks, prediction, sulfur-containin aromatic compounds, aero- , . bic microbiode r adation multiple linear re ression . , , , , , . , * ( Pr oj ect of Supported by Na- . , tional Natural Science Foundation of Ch ina ) : : , 25 , : 1997-06-13 38