摘要
针对高维数据集的复杂性,提出基于弹性网的两阶段模型平均方法,并将其应用于上证180指数的分析与预测研究中.首先通过弹性网进行变量降维并构建稀疏的候选模型;再根据Jackknife模型平均方法平均候选模型,最大限度用最少的成本获取更多的信息,减少有用信息的损失以提高模型预测精度,并使用各类预测误差指标来验证各预测模型的有效性.研究表明,两阶段模型平均方法可以有效降低上证180指数预测模型的预测误差;弹性网-JMA方法在高维有效样本下具有更好的预测表现和稳健性.
In terms of the complexity of the high dimensional data sets, is this paper, the two-stage model averaging method is proposed based on elsatic net and applies it to the research of analyzing and forecasting SSE 180 Index. Firstly, the dimension of the variables is reduced through elastic net and the sparse candidate models is constructed, and then the candidate models is averaged according to Jackknife model averaging method to gain more information with maximum limit and least cost and to reduce the loss of useful information so as to improve the forecast accuracy of the model. Finally, various forecasting error indicators will be applied to verify the effectiveness of the prediction models. Research shows that the two-stage model averaging method can reduce the forecasting error of the SSE 180 Index prediction models and the EN-JMA method shows better forecasting performance and robustness under the high dimensional valid samples.
作者
魏巍
王星惠
陈晓星
Wei Wei;Wang Xinghui;Cheng Xiaoxing(Anhui University)
出处
《哈尔滨师范大学自然科学学报》
CAS
2022年第6期47-53,共7页
Natural Science Journal of Harbin Normal University
基金
国家自然科学基金项目(11701005)
中国博士后科学基金面上资助(2019M662146)
安徽省社会科学规划项目(AHSKQ2020D63)
作者简介
通讯作者:王星惠
