摘要
将连接树的方法应用到吡啶类化合物 p Ka 值的预测中 ,并将该方法计算出的参数与量子化学参数相结合 .变量的两两相关性检验结果表明 ,所选择的 1 0个参数相关性较小 ,同时用交叉验证方法所得到的结果表明 ,所构造的多元回归模型十分稳定 .通过人工神经网法对回归模型进一步优化 ,得到满意的结果 .
For the prediction of pK_a of pyridine derivatives,the combinations of the parameters by using the method of tree structured fingerprint and the quantum-chemical parameters was performed. The results of the cross-correlation among the selected ten parameters showed that they possessed low correlations. And the results of the cross-validation method indicated that the multiple linear regression model was very steady. Furthermore,the pK_a values were estimated by using artificial neural networks. [WT5HZ]
出处
《高等学校化学学报》
SCIE
EI
CAS
CSCD
北大核心
2004年第6期1100-1103,共4页
Chemical Journal of Chinese Universities
基金
国家自然科学基金 (批准号 :2 0 0 770 2 6)资助
关键词
定量构效关系
连接树方法
吡啶类化合物
PKA值
人工神经网
Quantitative structure-property relationship
Tree structured fingerprint
Pyridine derivatives
pK_a value
Artificial neural networks
