摘要
本文聚焦2011-2021年上市公司在融资性贸易背景下与第三方协同财务造假的行为,基于五维度财务造假识别框架,运用机器学习方法,挖掘出针对识别财务造假至关重要的财务与非财务指标,并以*ST凯乐为例,实际验证了识别指标的有效性。研究发现,LightGBM模型在检测第三方协同财务造假方面表现优异,预付账款占总资产比例与投资活动水平异常是识别造假的关键线索;此外,供应商稳定性和采购集中度等非财务指标可以显著赋能财务造假识别。本研究推进了财务造假预测模型的理论发展,也拓展了机器学习在财务领域的应用,为资本市场监管提供了新的视角和有力工具。
Focusing on the 2011-2021 listed companies'collaborative financial fraud behavior with the third-party in the context of financing trade,based on the five dimensions financial fraud identification framework,using machine learning technology,this paper digs out the crucial financial and non-financial indicators of financial fraud identification,and takes*ST as an example to practically verifies the effectiveness of the identification index.It finds that the LightGBM model performs excellently in detecting third-party collaborative financial fraud,and the abnormal relationship between the proportion of prepayment in total assets with the investment activities level is a key clue to identify the fraud.In addition,nonfinancial indicators such as supplier stability and procurement concentration can significantly facilitate the financial fraud identification This study promotes the theoretical development of the financial fraud prediction model,and also expands the application of machine learning in the financial field,and provides a new perspective and a powerful tool for the capital market regulation.
出处
《中国注册会计师》
北大核心
2024年第10期63-74,5,共13页
The Chinese Certified Public Accountant
