论文部分内容阅读
通过构建分子顶点原子之间的关系对35个醛酮类化合物进行了结构表征,采用多元线性回归和逐步回归的方法建立了该类化合物结构与色谱保留指数之间的关系模型,模型的复相关系数(R)为0.978,标准偏差(SD)为37.305。采用“留一法”交互检验和外部样本预测的方法对模型的稳定性和预测能力进行了评价,交互检验的复相关系数(R_(cv))为0.970,标准偏差(SD_(cv))为43.477;外部预测的复相关系数(R_(test))为0.996,标准偏差(SD_(test)为17.004。结果表明所采用的分子结构描述符能够恰当表现该类化合物结构特征,所建模型具有良好的稳定性和预测能力。
The structures of 35 aldehydes and ketones were characterized by constructing the relationship between the molecular vertex atoms. The relationship between the structure and the chromatographic retention index was established by multivariate linear regression and stepwise regression. The model was correlated The coefficient (R) was 0.978 and the standard deviation (SD) was 37.305. The stability and predictive ability of the model were evaluated by the method of “leave one method” and the external sample prediction. The correlation coefficient (R cv) of the interaction test was 0.970 and the standard deviation (SD cv) ) Was 43.477, the correlation coefficient (R test) of external prediction was 0.996, and the standard deviation (SD test) was 17.004.The results showed that the molecular structure descriptors used in this study could properly represent the structural characteristics of the compounds, Has good stability and predictability.