摘要
针对现实中舞弊样本与非舞弊样本存在的数量不平衡情况,通过25个财务指标与2个非财务指标,运用过采样、欠采样技术及XGBoost模型进行财务报表舞弊识别研究。结果表明,SMOTE过采样方法与XGBoost模型的结合在非平衡数据集下具有较好的整体识别效果,对上市公司财务报表舞弊的智能识别有一定参考意义。
In view of the unbalance in the number of fraud samples and non-fraud samples in reality,a study on financial statement fraud identification is conducted by applying over-sampling,under-sampling techniques and XGBoost model to 25 financial indicators and 2 non-financial indicators.The results show that the combination of SMOTE over-sampling method and XGBoost model has a good overall identification effect in the unbalanced dataset,which has certain reference significance for the intelligent identification of financial statement fraud of listed companies.
作者
王琦
熊莎丽娜
詹柔
张露
杨鑫
张健
Wang Qi;Xiong Shaina;Zhan Rou;Zhang Lu;Yang Xin;Zhang Jian(School of Mathematics and Science,Southwest Forestry University,Kunming,Yunnan 650224,China)
出处
《计算机时代》
2023年第12期59-63,共5页
Computer Era
基金
云南省教育厅科学研究基金项目“基于非线性逻辑回归的M-Score模型优化研究”(2022J0523)
云南省高等学校大学生创新创业训练计划项目“基于数据挖掘的企业财务报表舞弊识别研究”。