基于二分类Logistic回归分析的桃仁等级预测研究

目的基于"成分反映活性,活性指向功效"的中药质量控制研究思路,建立用于桃仁饮片等级评价的Logistic模型。方法采用二元Logistic算法将桃仁饮片苦杏仁苷含量、体外凝血酶时间(TT)、ABTS清除率、DPPH清除率和羟自由基清除能力等指标进行关联分析,建立用于桃仁饮片的等级预测模型,并进行验证。结果 Logistic预测模型概率公式将18个批次桃仁饮片最终分为以内蒙桃仁为代表的优级、以甘肃桃仁为代表的良级、以步长桃仁为代表的中级和以陕西桃仁为代表的差级4个等级,各批次所对应的预测归属等级明确,拟合概率P值均大于98%。结论基于Logistic算法并结合"质控成分-体外效价"所建立的分级标准可用于桃仁饮片的等级分类,且结果准确可靠。

Research on grade prediction of Persicae Semen based on binary Logistic regression analysis

Objective A Logistic model for quality evaluation of Persicae Semen slices was constructed and its feasibility was verified in this study based on the thoughts on quality control of Chinese materia medica "Components reflect activity and activity points to efficacy". Methods Content of amygdalin, thrombin time (TT), ABTS clearance rate, DPPH inhibition rate, and hydroxyl radical scavenging ability in vitro were analyzed by a binary Logistic algorithm. Besides, a grade prediction model for Persicae Semen was established and verified. Results A total of 18 batches of Persicae Semen were divided into four grades:excellent (represented by Neimeng Persicae Semen), good (represented by Gansu Persicae Semen), medium (represented by Liang Persicae Semen), and poor (represented by Shaanxi Persicae Semen) according to the probability formula of the Logisitc model. All batches of Persicae Semen slices were classified accurately, manifested by the high value of fitting probability (P>98%). Conclusion The classification standard based on the Logistic algorithm involving quality control component-in vitro titer is applicable to Persicae Semen slices on the market. Classification results are accurate and reliable.