论文部分内容阅读
重点讨论对提高磷酰脲类除草剂分子活性预报准确性有明显影响的两方面因素──结构参数的选择和计算方法的应用．在选择结构参数时兼顾分子的局部活性基团效应和分子的整体性质，并运用了最优化计算方法──人工神经网络方法（ANN）．所得预报结果与实验值吻合较好（预测均方误差msc＝0．073），远优于多元回归所得的结果（msc＝0．82）．
It focuses on two factors that have a significant impact on improving the prediction accuracy of the molecular activity of the phosphonylurea herbicide - the choice of structural parameters and the application of calculation methods. Taking into account the local active group effect of the molecule and the global properties of the molecule in the selection of structural parameters, an optimization calculation method - Artificial Neural Network (ANN) is used. The predicted results are in good agreement with the experimental values (mean square error of prediction msc = 0.073), which is much better than multivariate regression (msc = 0.82).