基于改进随机森林算法的财务风险预警模型构建研究  

Research on the construction of financial risk early warning model based on improved randomforest algorithm

作　　者：颜铤 Yan Ting(Internet Information College of Chongqing Vocational College of Applied Technology,Chongqing Hechuan,401520,China)

机构地区：[1]重庆应用技术职业学院互联网信息学院,重庆合川401520

出　　处：《现代科学仪器》2023年第1期20-24,共5页Modern Scientific Instruments

基　　金：2022年重庆市教委科学技术研究项目,项目名称:《曼哈顿非负矩阵算法在人工神经网络的应用研究》,项目编号为KJQN202205302。

摘　　要：财务危机是影响企业发展的重要因素,也是企业在发展中不可避免的危险因素,因此对企业财务危机进行有效预警具有重要意义。为了实现企业财务风险预测和预警,研究提出了基于K折交叉验证的随机森林算法,同时结合时间序列分析来实现更为全面的预警。结果中显示,K折交叉验证随机森林算法的指标分类精度达到了0.907,并且研究构建的财务风险预警模型在ROC曲线的性能分析中表现出较高的线下面积,即具有较高的预测性能。以上结果表明,采用K折交叉验证改进随机森林算法能够较大程度上提升预警模型的财务指标分类效果,提升预警模型的综合预警能力,为我国企业发展提供财务风险规避策略。Financial crisis is an important factor affecting the development of enterprises,and it is also an inevitable risk factor in the development of enterprises.In order to realize the financial risk prediction and early warning of enterprises,a random forest algorithm based on k-fold cross validation is proposed,and a more comprehensive early warning is realized by combining time series analysis.The results show that the index classification accuracy of k-fold cross validation random forest algorithm reaches 0.907,and the financial risk early-warning model constructed by the research shows a high offline area in the performance analysis of ROC curve,that is,it has a high prediction performance.The above results show that the k-fold cross validation improved random forest algorithm can greatly improve the classification effect of financial indicators of the early warning model,improve the comprehensive early warning ability of the early warning model,and provide financial risk avoidance strategies for the development of Chinese enterprises.

关 键 词：财务危机 风险预警 K折交叉验证 随机森林 时间序列 

分 类 号：F832[经济管理—金融学]