摘要
文章引入机器学习算法,基于改进的带有自适应白噪声的完全集合经验模态分解(ICEEMDAN)-帝国竞争算法(ICA)—极限学习机(ELM),构建中国采购经理人指数预测模型,引入Diebold-Mariano统计量进行预测结果的比较。结果发现:ICEEMDAN技术可以准确提取数据中的有效信息,改进模型拟合效果;提出的组合模型ICEEMDAN-ICA-ELM预测效果优良,泛化能力强,误差较小,能够为PMI的走势提供新的预测方法。
This paper introduces machine learning algorithm,and relies on the improved complete ensemble empirical mode decomposition with adaptive noise(ICEEMDAN)-imperial competition algorithm(ICA)-extreme learning machine(ELM)to establish Chinese purchasing managers index prediction model,introducing Diebold-Mariano statistical magnitude to compare the forecasted results.The results show that the ICEEMDAN technology can accurately extract the effective information from data,and improve the model’s fitting effect;the proposed combined ICEEMDAN-ICA-ELM model has good prediction effect,strong generalization ability and small errors,which can provide a new way to predict the trend of PMI.
作者
相瑞兵
石亚男
马晓君
Xiang Ruibing;Shi Yanan;Ma Xiaojun(School of Statistics,Dongbei University of Finance and Economics,Dalian Liaoning 116025,China)
出处
《统计与决策》
CSSCI
北大核心
2020年第3期27-32,共6页
Statistics & Decision
基金
国家社会科学基金一般项目(19BTJ054)
国家自然科学基金资助项目(71772113
11701071)
教育部人文社会科学研究资助项目(18YJC910013)。
