基于长短期记忆神经网络的企业财务风险预警模型研究  

Early warning model of enterprise financial risk based on LSTM neural network

作　　者：林丹楠 李珊珊 肖世龙[3] 张德育[4] Lin Dannan;Li Shanshan;Xiao Shilong;Zhang Deyu(School of Information Engineering,Fujian Business University,Fuzhou 350012,China;School of Finance and Accounting,Fujian Business University,Fuzhou 350012,China;School of Art and Design,Shenyang Ligong University,Shenyang 110159,China;School of Information Science and Engineering,Shenyang Ligong University,Shenyang 110159,China)

机构地区：[1]福建商学院信息工程学院,福建福州350012 [2]福建商学院财务与会计学院,福建福州350012 [3]沈阳理工大学艺术设计学院,辽宁沈阳110159 [4]沈阳理工大学信息科学与工程学院,辽宁沈阳110159

出　　处：《南京理工大学学报》2021年第3期361-365,374,共6页Journal of Nanjing University of Science and Technology

基　　金：教育部产学合作协同育人项目(201802284028);福建省中青年教师教育科研项目(科技类)(JAT200394);辽宁省自然科学基金(20180520038)。

摘　　要：为了提高企业财务风险预警的准确度,采用长短期记忆(LSTM)神经网络算法建立企业财务风险预警模型。首先分析企业财务风险预警指标,并选取重要度高的指标构建企业财务风险预警特征样本。然后经过LSTM神经网络训练,采用遗忘门和记忆节点对历史财务数据进行遗忘和筛选,保留部分数据代入下次神经网络训练,通过反向传播获取最优权重与阈值。以企业财务风险预警准确度为目标函数,获得稳定的企业财务风险预警模型。实验证明,LSTM神经网络算法能够对企业财务重要指标进行预测,而且能正确设置企业财务风险预警阈值。通过和常用企业财务风险预警算法对比,本文算法的预警准确率更高。In order to improve the accuracy of enterprise financial risk early warning,long-short term memory(LSTM)neural network algorithm is used to establish an enterprise financial risk early warning model.Firstly,the financial risk early warning indicators of enterprises are analyzed,and the indicators with high importance are selected to construct the enterprise financial risk early warning characteristic samples.Then,the LSTM neural network is trained,the historical financial data is forgotten and screened by using the forgetting gate and memory node,and some data is reserved for the next neural network training to obtain the optimal weight and threshold through back propagation.Taking the accuracy of enterprise financial risk early warning as the objective function,a stable enterprise financial risk early warning model is obtained.Experimental results show that the LSTM neural network algorithm can predict the important financial indicators of enterprises,and can correctly set up enterprise financial risk early warning thresholds.Compared with the common enterprise financial risk early warning algorithm,the accuracy of this algorithm is higher.

关 键 词：长短期记忆 神经网络 企业财务风险预警 遗忘门 记忆节点 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]