论文部分内容阅读
对于19个1,1-二苯基乙烯衍生物,分别采用人工神经网络(网络结构为3-7-1)和线性回归分析方法,建立了其抗雌激素活性/C与扩展的引力指数Go、17号氢原子的净电荷Q和24号氧原子与17号氢原子间库仑力KL之间的QSAR模型,ANN模型的相关系数R=0.9999,标准偏差SD=3.05888E-4;MLR模型的相关系数R=0.9660,标准偏差SD=0.1010。结果表明人工神经网络是一种比较精密的拟合方法,具有良好的预测效果。
For 19 derivatives of 1,1-diphenylethylene, their anti-estrogen activity / C and extended gravitation index Go were established using artificial neural networks (network structure 3-7-1) and linear regression analysis , The net charge Q of 17th hydrogen atom and the QSAR model between 24th oxygen atom and Coulter force KL between 17th hydrogen atom. The correlation coefficient R of ANN model was 0.9999 and the standard deviation was SD = 3.05888E-4. The MLR model Correlation coefficient R = 0.9660, standard deviation SD = 0.1010. The results show that artificial neural network is a relatively accurate fitting method, with good predictive effect.