Forecasting financial distress of Chinese new energy listed companies using a novel hybrid model of Lasso-CCSM-VNWOA-GBDT

Effective forecasting the financial distress of energy enterprises can promote risk management and financial sustainability. Previous studies lack discussion of this topic in the emerging energy industry and ignore the sample imbalance problem and information screening. Therefore, the main work of this study is constructing a multi-dimensional evaluation system that integrates financial and non-financial indicators and then designing a hybrid forecasting model of Lasso-CCSM-VNWOA-GBDT. The model integrates the least absolute shrinkage and selection operator (Lasso) regression, cluster centroids (CC) algorithm, synthetic minority over-sampling technique (SMOTE), whale optimization algorithm improved by von Neumann topology (VNWOA), and gradient boosting decision tree (GBDT). The results show that, (1) the multi-dimensional evaluation system can comprehensively assess financial distress. In particular, the non-financial indicators of audit opinions, government subsidies, and number of R&D personnel are important identified variables. (2) The Lasso regression and CCSM algorithms are superior in indicator screening and sample balancing operations, the proposed model presents the advantage of forecasting financial distress for all evaluation criteria. (3) Total liabilities, operating income, and government subsidies are the top three marginal contributions to the forecasting of financial distress. This conclusion can help new energy enterprises improve risk warnings and achieve financially healthy and sustainable development.