新闻报道如何影响采购经理人指数——基于文本挖掘技术  

How News Reports Affect Purchasing Managers’Index——Based on Text Mining Techniques

作　　者：郑小强[1,2] 陈萌 ZHENG Xiaoqiang;CHEN Meng(School of Economics and Management,Southwest Petroleum University,610500,Chengdu,Sichuan,China;China Energy Index Research Center,Southwest Petroleum University,610500,Chengdu,Sichuan,China)

机构地区：[1]西南石油大学经济管理学院,四川成都610500 [2]西南石油大学中国能源指数研究中心,四川成都610500

出　　处：《特区经济》2023年第11期94-98,共5页Special Zone Economy

基　　金：西南石油大学杰出人才项目(编号:2018RW005)。

摘　　要：为减小媒体过度报道导致的供应链“牛鞭效应”对指数测算的影响,文章提出新的研究框架用于预测采购经理人指数。基于2017年1月至2022年3月的PMI相关新闻和指数数据,结合文本挖掘技术和SVM算法,提出分析预测模型,并对不同核函数分类效果进行对比分析,得到新闻关键词及分类效果较好的核函数。结果表明:多项式核函数的预测效果明显优于RBF、Sigmoid以及线性函数;为保证指标的全面性,引入多个指标评判,运用本文提出的框架对PMI进行分类预测,平均预测准确率超过75%,证明文章提出的理论框架具有一定的可靠性。在此基础上从提高舆论监控、推动多产业信息融合、促进信息产业更新迭代三个视角提出政策建议,减小媒体过度报道对指数测算的影响。In order to reduce the impact of the“bullwhip effect”on the index measurement caused by exces⁃sive media coverage,the paper proposes a new research framework for forecasting the purchasing managers'in⁃dex.Based on PMI-related news and index data from January 2017 to March 2022,the paper proposes an analyti⁃cal prediction model for PMI by combining text mining technology and SVM algorithm,and conducts a compara⁃tive analysis of the classification effects of different kernel functions to obtain news keywords and kernel functions with better classification effects.The research results show that the prediction effect of poly⁃nomial kernel function is significantly better than that of RBF,Sigmoid and linear function;to ensure the comprehensiveness of indicators,multiple indicators are introduced for judging,and the average prediction accuracy of PMI is over 75%by applying the framework proposed in this paper,which proves to a certain extent that the theoretical framework proposed in this paper has certain reliability.On this basis,policy sugges⁃tions are made from three perspectives:improving public opinion monitoring,promoting the integration of multi-industry information,and promoting the renewal and iteration of the information industry to reduce the influence of excessive media coverage on the index measurement.

关 键 词：采购经理人指数 文本挖掘 新闻报道 SVM算法 

分 类 号：F713.54[经济管理—市场营销]