挖掘年度信息披露报告中表外披露的关键信息,对于揭示企业财务困境具有重要的意义.本文选取了 2001年-2021 年我国 A 股 232 家被 ST 的上市公司和 232 家正常公司为研究对象,融合年度信息披露报告中的财务文本信息(TI)和财务指标(FI),构建 TI-FI-CatBoost 集成预测模型,用于预判上市公司的财务困境.研究结果表明:(1)TI-FI-CatBoost 模型集成分析了财务文本信息和财务指标,在财务危机预测研究中优于仅使用财务指标的CatBoost模型以及Logistic模型、SVM模型和XGBoost模型;(2)预测财务困境的最佳窗口期为被ST前两年;(3)年度信息披露报告中的财务文本信息对于预警财务困境具有重要的参考价值,中断、原告、锐减、亏损、明显、谨慎、未能、可能、风险、减持等词汇与财务危机显著正相关,成本、控制、规模、资本、缺陷、收益等词汇与财务危机显著负相关.
Integrated Prediction Method for Corporate Financial Distress Based on TI-FI-CatBoost
Exploring the key information disclosed off balance sheet in annual information dis-closure reports is of great significance for revealing the financial difficulties of enterprises.This arti-cle selects 232 ST listed companies and 232 normal companies in China's A-share market from 2001 to 2021 as the research objects.By integrating financial textual information(TI)and financial indica-tors(FI)from annual information disclosure reports,a TI-FI-CatBoost integrated prediction model is constructed to predict the financial difficulties of listed companies.The research results indicate that:(1)The TI-FI-CatBoost model integrates and analyzes financial text information and financial indica-tors,and is superior to the CatBoost model that only uses financial indicators,as well as the Logistic model,SVM model,and XGBoost model in financial crisis prediction research;(2)The best window period for predicting financial distress is two years before being ST;(3)The financial text infor-mation in the annual information disclosure report has important reference value for early warning of financial distress.Words such as interruption,plaintiff,sharp reduction,loss,obvious,cautious,failure,possibility,risk,reduction are significantly positively correlated with financial crisis,while words such as cost,control,scale,capital,defects,and returns are significantly negatively correlated with financial crisis.
Annual information disclosure reportST CompanyFinancial crisisTI-FI-CatBoost model
万方数据
