基于局部加权回归实现五味子醇提结果预测  

The Prediction of Fructus Schisandrae Alcohol Extraction Results Based on Local Weighted Regression Realization

作　　者：杨春静[1] 倪健[1] 游龙泰 冷新[1] 王文平[1] 董晓旭[1] 张欣[1] 杜雪莹[1] 尹兴斌[1] 

机构地区：[1]北京中医药大学中药学院,北京100102

出　　处：《现代中药研究与实践》2017年第6期44-47,共4页Research and Practice on Chinese Medicines

基　　金：国家重大新药创制(2014ZX09304306)

摘　　要：目的利用局部加权回归算法对五味子中五味子醇甲在不同提取因素水平下的提取结果进行预测。方法采用高效液相色谱法,对其进行方法学考察;测定不同提取时间和溶剂倍量下五味子醇提液中五味子醇甲的浓度,采用局部加权回归算法,基于上述数据和WEKA分析平台实现五味子甲提取结果的预测,对预测结果进行验证。结果基于39组实验数据,建立的回归最佳模型参数设置结果:kNN=9;Classifier:Linear Regression function;Nearest Neighbor Search Algorithm:Linear kNN search;Weigheted Kernel:Tricube.预测值和真实值之间的相对偏差基本控制在10%之内,验证结果良好。结论通过建立局部加权回归能良好的对五味子醇提取结果进行预测。Objective To establish the process of alcohol extraction of Schisandra chinensis based on the locally weighted regression algorithm. Methods High-performance liquid chromatography is used to investigate its methodology. And the method is used to determine concentration of Schisandrol A in different extraction time （t） and solvent load （M）. The process of alcohol extraction of Schisandra chinensis is established based on the above data and WEKA analysis platform by using locally weighted regression algorithm. Results Based on the 39 groups experimental data, the best parameter of the model are ： kNN = 9; Classifier： Linear Regression function; Nearest Neighbor Search Algorithm： Linear kNN search; Weigheted Kernel： Tricube. The relative deviation between the predicted values and real values basic control within 10 %, the verification results is good. Conclusion The model is applicative, which could forecast the dynamical process of fructus schisandrae alcohol extraction.

关 键 词：五味子 五味子醇甲 高效液相色谱法 局部加权回归 提取结果预测 

分 类 号：R283[医药卫生—中药学]