多次交叉验证对PLSDA模型的影响研究

发布时间：2018-05-18 21:11

本文选题：交叉验证 + PLSDA　；参考：《中国卫生统计》2017年01期

【摘要】：目的比较一次交叉验证和多次交叉验证对PLSDA最优模型的影响,探讨在个体正确分组和少数个体错分时,多次交叉验证对PLSDA最优模型稳定性的影响。方法打乱数据集中个体顺序进行多次交叉验证,通过一次交叉验证和多次交叉验证的方法对模拟数据和真实数据进行分析,使用成分数和MSEP等参数值来评价模型变异性和稳定性。结果模拟数据结果,仅进行1次交叉验证结果成分数为3,MSEP值为0.3792;在不打乱数据标签时,5000次交叉验证结果中,成分数范围是2~6,MSEP值的范围0.2569~0.5794;打乱5%的标签时,5000次交叉验证结果中,成分数范围是1~8,MSEP值的范围0.2061~0.6463;真实数据结果,进行1次交叉验证结果成分数为4,MSEP值为0.1376;10000次交叉验证成分数范围是4~10,MSEP范围是0.0802~0.3761。结论一次交叉验证结果不稳定,在应用PLSDA建模时,多次交叉验证在少量个体错分时能够获得稳定模型,建议使用多次交叉验证确保PLSDA模型稳定性。
[Abstract]:Aim to compare the effects of one cross validation and multiple cross validation on the stability of PLSDA optimal model, and to explore the effect of multiple cross validation on the stability of PLSDA optimal model when individuals are correctly grouped and a few individuals are misclassified. Methods the individuals in the data set were scrambled for multiple cross validation, and the simulated data and real data were analyzed by one cross validation and multiple cross validation. Parameters such as fraction and MSEP are used to evaluate the variability and stability of the model. Results the result of only one cross validation was 0.3792, and the score range was 0.2569 / 0.5794 when the data label was not tampered with 5000 times. The range of MSEP is 0.2061U 0.6463.The result of one cross validation is 0.137610000. The range of MSEP is 0.0802 / 0.3761. the result of one cross validation is 0.0802 / 0.3761.The result is as follows: (1) the range of MSEP is 0.2061 / 0.463.The result of one cross validation is 0.137610 000 times, and the range of MSEP is 0.0802 / 0.3761. Conclusion the results of one cross validation are unstable. When using PLSDA model, multiple cross validation can obtain a stable model when a few individuals misdivide. It is suggested that multiple cross validation should be used to ensure the stability of PLSDA model.
【作者单位】：哈尔滨医科大学公共卫生学院卫生统计学教研室;
【基金】：黑龙江省自然基金重点项目(ZD201314) 国家自然基金(81502889)
【分类号】：O212.1

【相似文献】