基于上市公司舆情网络的股票信息型操纵识别模型  

作　　者：张建锋[1] 

机构地区：[1]西安理工大学经济与管理学院

出　　处：《企业经济》2019年第5期147-154,共8页Enterprise Economy

基　　金：国家社会科学基金项目"股票操纵网络的演化机理与治理模式研究"(项目编号:16BGL066)

摘　　要：股票信息型操纵识别是证券监管部门处罚股票违规交易的主要依据,识别准确率是有效打击操纵行为的核心与关键。基于中国证监会2016年查处的股票信息型操纵案件处罚资料与数据,以上市公司舆情与传播媒体为节点构建了公司舆情网络。网络拓扑结构分析结果表明,公司舆情网络节点与连线数量在信息操纵时期明显较多,网络密度明显偏低。基于逐步逻辑回归构建了股票信息操纵识别模型,实证结果表明:网络密度与点度中心势是识别股票是否被实施信息操纵的主要指标,模型样本内检验准确率为87.14%。依此本文提出三点管理建议:监管资源应向主要监控指标倾向分配;根据证券市场环境动态调整识别模型参数;正确认识股票信息操纵双向效应,适度控制其负向作用。The recognition of stock information manipulation is the main basis for securities market supervision departments to punish illegal trading of stocks, and the recognition accuracy rate is the core and key to effectively strike the market manipulation. Based on the cases text and data of stock information manipulation published by Chinese securities regulatory commission in 2016, the paper constructs the company public opinion networks which take company public opinions and media as nodes. The analysis results of network topology show that the quantity of nods and connections of the networks are obviously more in the period of information manipulation, and the density of networks is obviously low. Based on the main network parameters selected by stepwise logistic regression, a stock information manipulation recognition model is constructed. The empirical results show that the network density and the degree centralization are the main indicators to recognize information manipulation. Through the in-sample test, the recognition accuracy of the model is 87.14%. Accordingly, this paper puts forward three management suggestions: the supervision resources should be allocated to main monitoring indicators;the model parameters should be adjusted dynamically according to the securities market environment;and the two-way effect of stock information manipulation shold be understood as to control moderately its negative role.

关 键 词：舆情网络 信息操纵 LOGIT模型 操纵识别 

分 类 号：F832.5[经济管理—金融学]