文章摘要
刘斌,董清浩.基于奇异谱分析的PMI组合预测模型[J].安徽建筑大学学报,2024,32():
基于奇异谱分析的PMI组合预测模型
PMI portfolio forecasting model based on singular spectrum analysis
投稿时间:2023-05-10  修订日期:2023-06-10
DOI:
中文关键词: PMI  奇异谱分析  SARIMA  支持向量回归
英文关键词: pmi  ssa  sarima  svr
基金项目:安徽省高等学校科学研究重点项目(2022AH050247);安徽建筑大学科研项目(2016QD118)
作者单位邮编
刘斌 安徽建筑大学数理学院 230601
董清浩* 安徽建筑大学数理学院 230601
摘要点击次数: 290
全文下载次数: 93
中文摘要:
      摘要:为了提高制造业采购经理人指数(PMI)的预测精度,结合PMI周期性和非线性的特点,提出了融合奇异谱分析(SSA)、季节差分自回归移动平均(SARIMA)模型和支持向量回归(SVR)模型的组合预测模型。该模型采用SSA将PMI分解为主要成分和噪声成分,借助SARIMA模型处理线性问题以及SVR模型处理非线性问题的优势,分别对两个成分建立相应的预测模型,针对主要成分选取SARIMA模型和SVR模型进行建模,噪声成分选取SVR模型进行建模,最后将各自得到的结果组合为最终的预测结果。实验显示:SSA-SARIMA-SVR模型的误差评价指标最低,预测效果最好,显示该模型对PMI走势的预测具有一定的参考价值。
英文摘要:
      Abstract:To improve the accuracy of forecasting the manufacturing purchasing managers"" index (PMI), a model combining singular spectrum analysis (SSA), seasonal difference autoregressive moving average model (SARIMA) and support vector regression (SVR) is proposed, with the characteristics of PMI such as cyclicality and non-linearity being taken into account. The model uses SSA to decompose PMI into major components and noisy components. Based on the advantages of the SARIMA model in dealing with linear problems and the SVR model in dealing with non-linear problems, the corresponding prediction models are established for each component separately, with the SARIMA model and the SVR model selected respectively for modelling the major components and the SVR model for modelling the noisy components. The final prediction results are combined as the final prediction results. The experimental results show that the SSA-SARIMA-SVR model has the lowest prediction evaluation index and the best prediction effect, indicating that the model has potential applications in predicting tendency of PMI.
查看全文   查看/发表评论  下载PDF阅读器
关闭

分享按钮