融合模型树的偏最小二乘法的中医药分析研究

发布时间：2018-04-16 16:32

本文选题：偏最小二乘法 + 非线性　；参考：《中国新药杂志》2017年17期

【摘要】：中医方药呈现多靶点、多成分、多药效指标等特点,决定了中医药数据具有多自变量、多因变量和非线性的特征。偏最小二乘法(PLS)以其内部交叉核验的本质,难以满足中医药非线性的特性,而模型树在回归建模时,由多个多元线性片段组成,对非线性数据有良好的拟合效果。基于此,本文提出了一种融合模型树的PLS。PLS外模型中的主成分仍按照原来的方法不断提取并累计t=(t_1,t_2,t_3,…),将这些主成分分别与原始被解释变量不断构建模型树,直到满足精度条件为止。分别在麻杏石甘汤君药平喘实验、止咳实验和UCI机器学习数据集上进行实验,结果表明,融合模型树的PLS对中医药数据有很好的适应性。
[Abstract]:The characteristics of traditional Chinese medicine (TCM) such as multi-target, multi-component and multi-efficacy index determine that TCM data have the characteristics of multi-independent variable, multi-dependent variable and nonlinear.Partial least square method (PLS) is difficult to satisfy the nonlinear characteristics of TCM because of its internal cross verification. In regression modeling, the model tree is composed of multiple linear segments, which has a good fitting effect on nonlinear data.Based on this, this paper proposes that the principal components of an external PLS.PLS model fused with the model tree are still extracted and accumulated according to the original method.The principal components and the original interpreted variables are used to construct the model tree until the precision condition is satisfied.The experiments were carried out on the experimental data sets of Maxingshi Gantangjun medicine for relieving asthma, cough relief and UCI machine learning. The results showed that the PLS of the fusion model tree had a good adaptability to the traditional Chinese medicine data.
【作者单位】：江西中医药大学计算机学院;
【基金】：国家自然科学基金项目(61363042);国家自然科学基金项目(61562045) 江西省自然科学基金重大项目(20152ACB20007) 江西中医药大学校级研究生创新专项资金项目(JZYC16S05)
【分类号】：O212;R28

【相似文献】