论文部分内容阅读
本文中建立了几个定量的模型预测80个烷基苯的沸点和79个烷基苯的摩尔体积。每个烷基苯的结构用其分子式得到的6个数字编码来描述。把这6个数字编码作为描述符,运用多元线性回归,多元非线性回归和人工神经网络地方法来分别建立定量构效关系模型。模型具有很好的预测性。沸点的3个预测模型,RMS偏差都小于9℃,摩尔体积的3个预测模型的RMS偏差都小于6 cm~3·mol~(-1)。
In this paper, several quantitative models were established to predict the boiling points of 80 alkylbenzenes and the molar volumes of 79 alkylbenzenes. The structure of each alkylbenzene is described by the six-digit code obtained from its formula. Using these six numerical codes as descriptors, the quantitative structure-activity relationship model was established respectively using multivariate linear regression, multivariate nonlinear regression and artificial neural network. The model is very predictive. The three prediction models of the boiling point have RMS deviations less than 9 ℃. The RMS deviations of the three prediction models for the molar volume are all less than 6 cm -3 mol -1.