摘要
应用多元线性回归分析和新近发展起来的人工神经网络方法进行了一类重要环境污染物多氯酚的定量构效关系研究，并用所建立的模型进行毒性预报。计算值与实验值的比较表明。前者的相关系数约为０．９２，后者的相关系数约为０．９９；后者的百分误差也明显小于前者；后者的预报能力略好于前者。文中还讨论了后者优于前者的原因。
Two quantitative structure-activity relationship (QSAR) models for chlorophenol are constructed by use of multivariate linear regresion analysis and artificial neural networks. They are used to predict toxicities of the chlorophenol not included in the training set. By comparing the calculated values and the experiment results, it was concluded that artificial neural networks approach for QSAR study of environmental pollutants is better than multivariate linear regression analysis.
出处
《环境科学学报》
CAS
CSSCI
CSCD
北大核心
1996年第2期190-194,共5页
Acta Scientiae Circumstantiae
基金
国家自然科学基金
