论文部分内容阅读
在多氯二苯-p-二氧杂芑(PCDD)类化合物极性气相色谱相对保留因子的QSPR分析中采用岭回归处理,取岭参数k=0.02时可有效抑制量子化学参数间多重相关性的不良影响。与传统的逐步回归法的结果相比,得到的QSPR函数表达更加符合微观过程的物理、化学原理。在外推区域对未知化合物性质的预报精度有明显的提高。证实岭回归方法得到的QSPR模型更为合理并具有更高的稳定性。
In the QSPR analysis of the relative retention factor of polar gas chromatography for polychlorinated dibenzo-p-dioxins (PCDD) compounds, Ridge Regression was adopted. When ridge parameter k = 0.02, the multiple correlations between quantum chemical parameters The adverse effects. Compared with the traditional stepwise regression method, the QSPR function expression is more in line with the physical and chemical principles of microscopic processes. In the extrapolation region, the prediction accuracy of the properties of unknown compounds is obviously improved. Confirm that the QSPR model obtained by ridge regression method is more reasonable and has higher stability.