基于异质特征融合的财务风险预测研究——以新能源上市公司为例

Research on Financial Risk Prediction Based on Heterogeneous Feature Fusion:Take New Energy Listed Companies as an Example

上市公司年报中的管理层讨论与分析(Management Discussion&Analysis,MD&A)包含了重要的非财务信息和前瞻性信息,股吧则反映了投资者对公司的态度,通过对MD&A和股吧文本信息的挖掘与分析可以提高财务风险的预测能力。课题组提出了基于异质特征融合的财务风险预测模型,模型选取风险区分能力显著的财务指标作为基本指标,融入年报中的管理层讨论与分析(MD&A)和东方财富股吧文本信息,并以Stacking和Bagging两种不同的决策融合方案进行风险预测。基于2022年中国151家新能源上市公司样本的实证结果表明:不同来源的金融文本信息构建的指标均可以提高财务风险模型的预测准确率,且融合多维度和多来源文本信息的指标融合方案整体预测性能最优,其在Bagging决策融合策略的预测准确率最高达到91.3%。

The management discussion&analysis(MD&A)in the annual report of listed companies contains important non-financial information and forward-looking information,and the stock bar reflects the attitude of investors towards the company.This paper proposes a financial risk prediction model based on the fusion of heterogeneous features,which selects financial indicators with significant risk discrimination ability as the basic indicators,integrates the management discussion and analysis(MD&A)and the text information of Oriental Fortune Stock Bar in the annual report,and uses two different decision fusion schemes of Stacking and Bagging for risk prediction.The empirical results based on the sample of 151 new energy listed companies in China in 2022 show that the indicators constructed from different sources of financial text information can improve the prediction accuracy of the financial risk model,and the overall prediction performance of the index fusion scheme integrating multi-dimensional and multi-source text information is the best,and the prediction accuracy of the Bagging decision fusion strategy is up to 91.3%.