• 关键词：

• 基金项目：国家自然科学基金

• 王桂莲

• 白乃彬

• 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。

• Abstract：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.

• Key words：

• 
  摘要点击次数： 1665 全文下载次数： 2305