基于非平衡数据处理和多变量筛选方法的上市公司财务困境预测研究

Financial Distress Prediction of Listed Companies Based on Unbalanced Data Processing and Multivariate Selection Methodology

在国内供给侧结构性改革背景下,市场环境复杂多变,公司债务违约频发,建立一种及时有效的财务困境预警模型十分必要.已有的多数困境预测模型尚未有效解决数据集不平衡、关键预测指标选取不稳定、样本匹配存在随机性等问题,且并不适应于当下中国复杂多变的市场状况.为此,本文采用Bootstrap重抽样方法构建1000个研究样本,通过LASSO(least absolute shrinkage and selection operator)变量选择技术筛选关键预测因子构建提前3年预测的Logit模型,并在预测阶段将样本进行1000次随机切割和预测以降低随机误差.结果表明,由Bootstrap样本组建方式结合LASSO构建的Logit困境预测模型相比传统应用的“同行业资产规模相近”方式所构建模型的预测能力更强.另外,该嵌入Bootstrap式LASSO-logit模型比主流的O-Score模型、ZChina-Score模型预测效果更好,准确率提高10%,更加适用于中国时变的市场.本文所构建模型能帮助公司利益相关者更好地识别财务困境并及时做出调整,以降低公司债券违约率或避免发生公司违约.

In the context of China’s domestic supply-side structural reforms,the market environment has become increasingly complex and volatile,accompanied by a high frequency of corporate defaults.It is necessary to establish a timely and effective financial distress prediction model.However,most existing distress prediction models have not adequately addressed critical issues such as the imbalance of datasets,the instability in selecting key predictive indicators,and the randomness in sample matching.Moreover,these models are not well-suited for this dynamic market in China.To address these limitations,this study initially employs the Bootstrap resampling method to construct 1000 research samples and uses the LASSO(least absolute shrinkage and selection operator)variable selection technique to screen key predictors.Then,this study constructs a logit model for predicting financial distress three years in advance using LASSO-selected key predictors.To mitigate the impact of random errors,the samples are further subjected to 1000 random splits and predictions.The results demonstrate that the LASSO-logit prediction model,constructed by combining the Bootstrap sample construction method with LASSO,exhibits superior predictive ability compared to the models constructed using the traditional“similar asset size in the same industry”approach.Furthermore,this LASSO-logit model outperforms mainstream O-Score and ZChina-Score models,with a 10% increase in accuracy,making it more suitable for China’s dynamic market.Overall,the model constructed in this study can help stakeholders of companies better identify financial distress and make timely adjustments to reduce corporate bond default rates or avoid corporate defaults.