Scholarlink

<p>To address the issue of “frequency mismatch” and “curse of dimensionality” in forecasting the Chinese Medicine Materials Price Index (CMMPI), we propose a Penalized Reverse Unrestricted Mixed Data Sampling (PRU-MIDAS) model. The model introduces the LASSO regularization mechanism within the Reverse Unrestricted Mixed Data Sampling (RU-MIDAS) framework to achieve variable selection and sparse parameter estimation for mixed-frequency data. This effectively addresses the challenges of analyzing and predicting high-frequency variables using low-frequency variables in big data environments. Our study takes the “Chinese Medicine Price Index: Composite 200” from September 1, 2014, to May 31, 2025 as the response variable, and integrates multidimensional monthly and daily explanatory variables covering supply, demand, macro-finance, foreign trade, and circulation to construct the forecasting system. Empirical results demonstrate that the PRU-MIDAS model delivers strong fitting performance and predictive accuracy, significantly outperforming traditional same-frequency models (such as RW, AR, and ARDL) as well as the RU-MIDAS model. Moreover, it is capable of revealing the real-time dynamic interrelationships among variables in detail. Our research not only provide a novel methodological support for forecasting price trends in the Chinese medicine market but also offers a scientific basis for government regulation and enterprise decision-making.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on the Prediction of Chinese Medicinal Materials Price Index Based on the PRU-MIDAS Model

  • 期刊论文

  • 发表日期：

    2025年11月20日

查看全文
  • Mengnan Xu,
  • Zhipei Feng,
  • Cheng Wang

摘要

To address the issue of “frequency mismatch” and “curse of dimensionality” in forecasting the Chinese Medicine Materials Price Index (CMMPI), we propose a Penalized Reverse Unrestricted Mixed Data Sampling (PRU-MIDAS) model. The model introduces the LASSO regularization mechanism within the Reverse Unrestricted Mixed Data Sampling (RU-MIDAS) framework to achieve variable selection and sparse parameter estimation for mixed-frequency data. This effectively addresses the challenges of analyzing and predicting high-frequency variables using low-frequency variables in big data environments. Our study takes the “Chinese Medicine Price Index: Composite 200” from September 1, 2014, to May 31, 2025 as the response variable, and integrates multidimensional monthly and daily explanatory variables covering supply, demand, macro-finance, foreign trade, and circulation to construct the forecasting system. Empirical results demonstrate that the PRU-MIDAS model delivers strong fitting performance and predictive accuracy, significantly outperforming traditional same-frequency models (such as RW, AR, and ARDL) as well as the RU-MIDAS model. Moreover, it is capable of revealing the real-time dynamic interrelationships among variables in detail. Our research not only provide a novel methodological support for forecasting price trends in the Chinese medicine market but also offers a scientific basis for government regulation and enterprise decision-making.