摘要
上市公司财务造假是一种违规行为,对其精确、有依据的预测有一定的研究价值.使用多种统计方法提取特征,并结合经济学意义得到了最终特征集.针对数据的不平衡问题,采用过采样、欠采样以及SMOTE采样等方法进行处理.最后采用3种神经网络模型融合的方法,以AUC值为评价指标来预测财务造假的上市公司.
Financial fraud is violation,and its accuracy and valid prediction has certain research values.A variety of statistical methods are used to extract features,and the final feature set is obtained combining with economic significance.Aiming at the imbalance of data,over sampling,under sampling and smote sampling are used to deal with it.Finally,three neural network model fusions are used to predict the listed companies with financial fraud with AUC value as the evaluation index.
作者
蔡景波
蔡志杰
CAI Jingbo;CAI Zhijie(Shenzhen Digquant Network Technology Co.,Ltd.,Shenzhen,Guangdong 518000,China;School of Mathematical Sciences,Fudan University,Shanghai 200433,China;Shanghai Key Laboratory for Contemporary Applied Mathematics,Shanghai 200433,China;Key Laboratory of Nonlinear Mathematical Models and Methods of Ministry Education,Shanghai 200433,China)
出处
《数学建模及其应用》
2021年第3期54-59,共6页
Mathematical Modeling and Its Applications
关键词
特征选择
深度学习
神经网络
不平衡处理
财务造假
feature selection
deep learning
neural network
imbalance treatment
financial fraud
