融合模型树的偏最小二乘法的中医药分析研究
|
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。
Traditional Chinese medicines( TCM) present features of more compositions,more targets and more efficacies. Therefore, the collected data of TCM exist multi-components, multi-targets and nonlinear characteristics. Partial least square( PLS) can't adapt to the characteristics of the TCM data due to its own nonlinear regression. However,model tree( MT),which is made up of many multiple linear segments,has a good fitting effect to nonlinear data. Based on this,a new method combining PLS and MT to analysis and predict the data is proposed,employ native PLS method to extract main ingredients continually and accumulate it,then build Model Tree through the main ingredients and the original explanatory variables one step by step,until the precision requirements are met. Using the data of the maxingshigan decoction of the monarch drug to treat the asthma or cough and five sample sets in the UCI machine learning repository,the experimental results showed that the PLS and model tree have good adaptability for the TCM data.
Traditional Chinese medicines( TCM) present features of more compositions,more targets and more efficacies. Therefore, the collected data of TCM exist multi-components, multi-targets and nonlinear characteristics. Partial least square( PLS) can't adapt to the characteristics of the TCM data due to its own nonlinear regression. However,model tree( MT),which is made up of many multiple linear segments,has a good fitting effect to nonlinear data. Based on this,a new method combining PLS and MT to analysis and predict the data is proposed,employ native PLS method to extract main ingredients continually and accumulate it,then build Model Tree through the main ingredients and the original explanatory variables one step by step,until the precision requirements are met. Using the data of the maxingshigan decoction of the monarch drug to treat the asthma or cough and five sample sets in the UCI machine learning repository,the experimental results showed that the PLS and model tree have good adaptability for the TCM data.
引文
[1]张伯礼,王永炎.组分配伍研制现代中药的理论与实践——方剂关键科学问题的基础研究[M].沈阳:辽宁科学技术出版社,2010.
[2]陆洪涛.偏最小二乘回归数学模型及其算法研究[D].北京:华北电力大学,2014.
[3]ABDI H,WILLIAMS L.Partial least squares methods:partial least squares correlation and partial least square regression[M].Humana Press,2013:549-579,930.
[4]王惠文,吴载斌,孟洁.偏最小二乘回归的线性与非线性方法[M].北京:国防工业出版社,2006.
[5]邵辉,李芳.基于树模型算法的动态网页信息抽取研究[C].2005.
[6]夏巧生.非线性偏最小二乘建模方法及在近红外光谱建模上的应用[J].计算机与应用化学,2014(1):109-112.
[7]朱志鹏,杜建强,余日庆,等.融入深度学习的偏最小二乘优化方法[J].计算机应用研究,2017,34(1):87-90.
[8]WOLD H.Nonlinear estimate by iterative least squares procedures[M]//Research Papers in Statistics.Wiley,New York,1966.
[9]QUINLAN JR.Learning with continuous classes[C].Singapore:Proceedings of the 5th Australian joint Conference on Artificial Intelligence,1992.
[10]QUINLAN JR.Learning with continuous classes[C].Singapore:Proceedings Australian Joint Conference on Artificial Intelligence.World Scientific,1992.
[11]HSIN CHU.UCI Machine Learning Repository[EB/OL].[2007-08-03].http://archive.ics.uci.edu/ml/datasets/Concrete+Compressive+Strength.
[12]PINAR TUFEKCI,HEYSEM KAYA.UCI Machine Learning Repository[EB/OL].[2014-03-26].http://archive.ics.uci.edu/ml/datasets/Combined+Cycle+Power+Plant.
[2]陆洪涛.偏最小二乘回归数学模型及其算法研究[D].北京:华北电力大学,2014.
[3]ABDI H,WILLIAMS L.Partial least squares methods:partial least squares correlation and partial least square regression[M].Humana Press,2013:549-579,930.
[4]王惠文,吴载斌,孟洁.偏最小二乘回归的线性与非线性方法[M].北京:国防工业出版社,2006.
[5]邵辉,李芳.基于树模型算法的动态网页信息抽取研究[C].2005.
[6]夏巧生.非线性偏最小二乘建模方法及在近红外光谱建模上的应用[J].计算机与应用化学,2014(1):109-112.
[7]朱志鹏,杜建强,余日庆,等.融入深度学习的偏最小二乘优化方法[J].计算机应用研究,2017,34(1):87-90.
[8]WOLD H.Nonlinear estimate by iterative least squares procedures[M]//Research Papers in Statistics.Wiley,New York,1966.
[9]QUINLAN JR.Learning with continuous classes[C].Singapore:Proceedings of the 5th Australian joint Conference on Artificial Intelligence,1992.
[10]QUINLAN JR.Learning with continuous classes[C].Singapore:Proceedings Australian Joint Conference on Artificial Intelligence.World Scientific,1992.
[11]HSIN CHU.UCI Machine Learning Repository[EB/OL].[2007-08-03].http://archive.ics.uci.edu/ml/datasets/Concrete+Compressive+Strength.
[12]PINAR TUFEKCI,HEYSEM KAYA.UCI Machine Learning Repository[EB/OL].[2014-03-26].http://archive.ics.uci.edu/ml/datasets/Combined+Cycle+Power+Plant.