论文部分内容阅读
为了研究人工神经网络预测有机化合物生物降解的性能,运用多元线性回归方法和误差反向传递人工神经网络模型以基团代码作为结构描述符,分别拟合、预测了一批含硫芳香族化合物的一级好氧生物降解速率常数.结果表明,由于神经网络自动考虑了基团间的交互作用,它对生物降解这类复杂问题有极高的求解能力,预测的均方误差为0.00102,远低于线性回归模型的预测误差0.01591
In order to study the performance of artificial neural network in predicting the biodegradability of organic compounds, using multivariate linear regression method and error propagation artificial neural network (ANN) model with group code as structure descriptor, a series of sulfur-containing aromatic compounds An aerobic biodegradation rate constant. The results show that neural network has high solvability for complex problems such as biodegradation due to the automatic interaction between groups. The mean square error of prediction is 0.00102, much lower than the prediction error of linear regression model 0.01591