基于偏最小二乘法(PLS)构建大型突发公共卫生事件创伤预后预测
|
摘要
目的采用汶川大地震后一组伤员救治医疗数据,以偏最小二乘(partial least square,PLS)算法为核心,探讨影响因素并构建一种新的大规模突发公共卫生事件创伤救治预后预测模型。方法以四川省医学科学院.四川省人民医院2008年5月收治的地震伤员数据库作为建模资料来源作为训练集,建模过程包括成分提取和建立回归方程,模型优度判别采用交叉检验。运算平台采用EXCEL和Matlab软件。结果通过相关性分析得到9个因素的相关性,VIP值得到5个影响结局的重要因素,并建立了大规模突发公共卫生事件创伤救治预后预测模型。结论本研究构建的以PLS算法为基础的模型,能可靠的帮助筛选大规模公共卫生事件中创伤患者预后预测的关键性影响因素以及综合评价指标。
Objective To illustrate a new method based on "partial least square"(PLS) algorithm to establish a predictive model which can be used for public health and trauma prognosis forecast,with bioinformatics data of clinical trauma services based on information initially gathered from the aftermath of the massive Wenchuan earthquake on May 12,2008.Methods The data come from the database of Sichuan Provincial People's Hospital.There are two steps in the modeling process: principal components analysis;and using these components to establish a linear regression equation.Cross validation is introduced for best fit of the PLS model.Computation is carried out using the EXCEL and Matlab.Results A total of 9 variation Correlation coefficient were obtained from Correlation Analysis,and obtain 5 important variation and establish a predictive model which can be used for public health andtrauma prognosis forecast.Conclusion The model based on PLS has a high degree of reliability to analyze important factors of prognosis for public health events for the purpose of predicting patient outcome.
Objective To illustrate a new method based on "partial least square"(PLS) algorithm to establish a predictive model which can be used for public health and trauma prognosis forecast,with bioinformatics data of clinical trauma services based on information initially gathered from the aftermath of the massive Wenchuan earthquake on May 12,2008.Methods The data come from the database of Sichuan Provincial People's Hospital.There are two steps in the modeling process: principal components analysis;and using these components to establish a linear regression equation.Cross validation is introduced for best fit of the PLS model.Computation is carried out using the EXCEL and Matlab.Results A total of 9 variation Correlation coefficient were obtained from Correlation Analysis,and obtain 5 important variation and establish a predictive model which can be used for public health andtrauma prognosis forecast.Conclusion The model based on PLS has a high degree of reliability to analyze important factors of prognosis for public health events for the purpose of predicting patient outcome.
引文
[1]王惠文.偏最小二乘回归方法及其应用[M].北京:国防工业出版社,1999,150～234
[2]Bookstein FL.An overview of partial least squraes from the viewpointof the natural science.Proceedings of the 6th international conferenceon partial least squares and related methods[M].Beijing,China:Pub-lishing house of electronics industry,2009,3～7
[3]Ruge M,Strohe HG.Analysis of the sentiments and expectations inthe german.Proceedings of the 6th international conference on partialleast squares and related methods[M].Beijing,China:Publishinghouse of electronics industry,2009,244～247
[4]Wold S,Sj str mM,Eriksson L.PLS-regression:a basic tool of chemo-metrics[J].Chemometr intell,2001,58:109～130
[5]Tenenhaus M.Application to the analysis of life time data.In:espositoPLS and related methods[J].Proceedings on the PLS’01 internationalsymposium,cisia,Pairs,2001,131～140
[6]Lorber A,Wangen LE,Kowalski,et al.A theoretical foundation forthe PLS algorithm[J].Chemometrics,1987,119
[7]Bastien P,Vinzi VE,Tenenhaus M.PLS generalized linear regression[J].Computational Statistics&Data Analysis,2005,48:17～46
[2]Bookstein FL.An overview of partial least squraes from the viewpointof the natural science.Proceedings of the 6th international conferenceon partial least squares and related methods[M].Beijing,China:Pub-lishing house of electronics industry,2009,3～7
[3]Ruge M,Strohe HG.Analysis of the sentiments and expectations inthe german.Proceedings of the 6th international conference on partialleast squares and related methods[M].Beijing,China:Publishinghouse of electronics industry,2009,244～247
[4]Wold S,Sj str mM,Eriksson L.PLS-regression:a basic tool of chemo-metrics[J].Chemometr intell,2001,58:109～130
[5]Tenenhaus M.Application to the analysis of life time data.In:espositoPLS and related methods[J].Proceedings on the PLS’01 internationalsymposium,cisia,Pairs,2001,131～140
[6]Lorber A,Wangen LE,Kowalski,et al.A theoretical foundation forthe PLS algorithm[J].Chemometrics,1987,119
[7]Bastien P,Vinzi VE,Tenenhaus M.PLS generalized linear regression[J].Computational Statistics&Data Analysis,2005,48:17～46
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心