论文部分内容阅读
通过非氢原子自身及非氢原子之间的关系对27个聚合物单体进行了结构表征,采用多元线性回归(MLR)和逐步回归(SMR)建立了聚合物结构与介电常数的关系(QSPR)模型,同时采用“留一法”对所得模型稳健性及预测能力进行了评价。模型复相关系数(R)为0.924,“留一法”交叉验证系数(Q2)为0.730。结果表明所采用的结构描述符可以恰当反映聚合物结构特征,构建的模型具有良好的稳健性和预测能力。
27 polymer monomers were characterized by the relationship between non-hydrogen atoms themselves and non-hydrogen atoms. The relationship between polymer structure and dielectric constant was established by multivariate linear regression (MLR) and stepwise regression (SMR) QSPR) model, and evaluated the robustness and predictive ability of the model by using “leave one method”. The model has a correlation coefficient (R) of 0.924 and the “leave one method” cross-validation coefficient (Q2) is 0.730. The results show that the structure descriptors used can properly reflect the structural characteristics of polymers, and the constructed models have good robustness and predictive ability.