摘要
本文采用简单的化学基团描述符来表征化学物质的结构、以底物的最大比去除速率表征生物可降解性的大小,运用自编的人工神经网络对芳香族化合物的生物可降解性进行了研究.试验过程中,按芳香族化合物最大比底物去除速率的大小将其生物可降解性分为四组:不降解、难降解、可降解、易降解.随机抽取10%的有机化合物作为预测集,余下的作为训练集,共进行了四次预测试验,正确预测率分别达80%,80%,4O%,80%,这显示了人工神经网络对芳香族化合物的生物可降解性有较好的预测性能.
Based on the substituted functional groups and their positions as the chemical structure descriptors, and the substrates' maximum specific biodegradation rates as the ability of their biodegran-dability. the aromatic compounds' degradabilities were studied using the authors constructed artificial nerual networks. During the experiment, the aromatic compounds' degradabilities were artificially classified as hardly biodegradable, hard biodegradable, biodegradable, and easy biodegradable according their maximum specific biodegradation rates, 10% substances were randomly subtracted as the prediction sets, and the rest were treated as the training sets. The correct prediction rates were 80% , 80%, 40%, and 80% respectively in the total four tests. These results showed that the artificial neural networks has the good ability to predict the biodegradability of aromatic substances.
出处
《环境化学》
CAS
CSCD
北大核心
2000年第1期48-52,共5页
Environmental Chemistry
关键词
神经网络
芳香族化合物
生物可降解性
污染物质
artificial neural networks, aromatic substances, biodegradability prediction.
