基于Lasso类方法在时间序列变量选择中的应用  

Application Research on Variable Selection in Time Series Based on a Class of Lasso Method

作　　者：杨丽娟[1] 马云艳[1] 

机构地区：[1]鲁东大学数学与统计科学学院,山东烟台264039

出　　处：《鲁东大学学报（自然科学版）》2016年第1期14-18,97,98,共7页Journal of Ludong University:Natural Science Edition

基　　金：鲁东大学博士人才引进基金(LY2013001;LY201222);鲁东大学名校工程精品课程项目(20130495;20130496);山东省科技发展计划项目(2012YD01056)

摘　　要：在时间序列场合下对Lasso方法与Adaptive Lasso方法进行比较,数值模拟结果表明Adaptive Lasso方法比Lasso方法更加有效.通过对沪深300指数的技术指标的历史数据进行变量选择,比较发现二者均可以有效并准确地选择出合适变量,且Adaptive Lasso方法的参数估计相对更加精确.最后,根据选出变量及其参数进行预测,结果表明Adaptive Lasso方法变量选择和参数估计效果良好.The Lasso method and Adaptive Lasso method were compared in time series. The simulation results indicate that Adaptive Lasso is more effective than the Lasso method. By using these two methods for variable selection in the HS300 index of the historical data,the results show that the two methods can select variable effectively and accurately,but parameters of Adaptive Lasso method are more accurately. Forecast results according to the selected variables and their parameters also show that the variable selection and parameter estimation of the Adaptive Lasso method are effective.

关 键 词：Lasso方法 变量选择 时间序列 

分 类 号：O212.1[理学—概率论与数理统计]