摘要
对财务危机预警模型进行研究,有助于企业及早发现可能的风险,制定相应对策加强财务管理,对完善资本市场体系,推动经济高质量发展意义重大.本研究提出了一种基于主成分分析(Principal Component Analysis,PCA)和卷积神经网络(Convolutional Neural Network,CNN)相融合的财务危机预警方法.首先,针对海量财务数据存在噪音的问题,构建表征企业财务风险的主成分分析框架,实现备选指标数据的预处理.其次,根据预警指标的特征,通过对财务预警数据的反向学习训练,构建轻量化的三层卷积神经网络模型,预测企业是否陷入财务危机.最后,与现有机器学习财务预警方法进行对比,新模型显示了较高的预测准确率.
Research on the financial crisis early warning model is helpful for enterprises to find the possible risks as early as possible and formulate corresponding countermeasures to strengthen financial management,which is of great significance to improve the capital market system and to promote the high-quality economic development.This research proposes a financial crisis early warning method based on Principal Component Analysis(PCA)and Convolutional Neural Network(CNN).Firstly,to mitigating the effects of the noise in massive financial data,the PCA framework is constructed to pre-process candidate indicator data and represent the financial risk of enterprises.Secondly,according to the characteristics of early-warning indicators,a lightweight three-layer CNN model is established to predict the financial crisis based on the opposition learning and training of financial early-warning data.Finally,compared with the existing machine learning financial early warning methods,the new model shows a higher prediction accuracy.
作者
谭媛元
陈建英
孙健
TAN Yuan-yuan;CHEN Jian-ying;SUN Jian(Finance Department, Southwest University, Chongqing 400715, China;School of Economics and Management, Southwest University, Chongqing 400715, China;School of Electronic and Information Engineering, Southwest University, Chongqing 400715, China)
出处
《西南师范大学学报（自然科学版）》
CAS
2021年第5期73-80,共8页
Journal of Southwest China Normal University(Natural Science Edition)
基金
重庆市社会科学规划项目(2020PY50,2018PY61)
中央高校基本科研业务费专项资金项目(SWU1809228).
