Financial Distress Prediction of Listed Companies Based on Unbalanced Data Processing and Multivariate Selection Methodology
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 predic-tion 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 mar-ket 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 predic-tions.The results demonstrate that the LASSO-logit prediction model,constructed by combining the Bootstrap sample construction method with LASSO,exhibits su-perior 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.
NETL
NSTL
万方数据
