基于数智技术的企业风险动态特征挖掘方法研究  

Enterprise risk dynamic feature mining method based on digital intelligence technology

作　　者：屈映池 杨梅 肖智[1] Qu Yingchi;Yang Mei;Xiao Zhi(School of Economics and Business Administration,Chongqing University,Chongqing 400044,China)

机构地区：[1]重庆大学经济与工商管理学院,重庆400044

出　　处：《数智技术研究与应用》2025年第1期50-60,共11页SmartTech Innovations

基　　金：国家自然科学基金(71671019、72071021)。

摘　　要：随着经济不确定性的增加,企业竞争日益加剧,企业风险具有非线性、快速变化和不确定性的特征,且不同企业应对风险能力各不相同。现有模型主要从线性或特有的非线性的视角出发,缺少对各个企业具体的风险非线性动态特征的刻画。因此,文章从数据驱动的视角出发,基于近期事件对企业风险影响更大的特点,构建了一个非线性集成动态隐马尔可夫-自组织神经网络(GMMHMM-SOM)模型,从微观层面挖掘企业风险动态特征。通过处理1753家上市公司的具有不同样本动态特征的数据,GMMHMM识别了企业风险状态的变化,得到风险状态转移概率矩阵。SOM挖掘风险状态转移概率矩阵,分析企业风险不稳定性的特征。研究发现,企业利润率和风险状态不稳定性共有六类特征并呈现“倒U形”的特点,平均利润率在区间[7.8%,8.08%]时企业风险最小。研究还发现,企业通过在利润率的基础上对风险进行进一步的分析,有助于企业采取差异化的管理措施应对企业风险变化,进而实现更好的经济效益。With the increase of economic uncertainty,the competition of enterprises is increasingly intensified,and the enterprise risk has the characteristics of nonlinear,rapid change and uncertainty.Different enterprises have different abilities to deal with the risk.Existing models mainly start from a linear or unique nonlinear perspective and lack the characterization of the nonlinear dynamic characteristics of each enterprise’s risk.Therefore,this paper constructs a nonlinear integrated dynamic Hidden MarkovSelfOrganizing Neural Network(GMMHMMSOM)model from a datadriven perspective and based on the characteristics that recent events have a greater impact on enterprise risk,to mine enterprise risk dynamics at the microlevel.By processing the data of 1753 listed companies with different dynamic characteristics of samples,GMMHMM identifies the change of enterprise risk state and obtains the risk state transition probability matrix.SOM mines the risk state transition probability matrix and analyzes the characteristics of enterprise risk instability.The study finds that there are six types of attributes of the fluctuation of enterprise profit rate and risk status,which show the characteristics of“inverted Ushaped”,and the enterprise risk is the smallest when the average profit rate is in the interval[7.8%,8.08%].The study found that enterprises can further analyze risks on the basis of the profit rate,which helps enterprises to take differentiated management measures to deal with changes in enterprise risks,thereby achieving better economic benefits.

关 键 词：数智技术 企业风险 利润率 隐马尔可夫链 自组织神经网络 

分 类 号：F27[经济管理—企业管理]