Multi-window weighted random forests for listed companies default prediction: a dual-criterion temporal framework with hybrid feature selection

Key Points

• The proposed model achieves superior accuracy in default prediction compared to traditional methods, highlighting its effectiveness.
• Key evidence from the model indicates significant feature importance, which enhances prediction capabilities over a dual-criterion framework.
• This study employs a hybrid feature selection method integrated within a weighted random forests approach, allowing for more precise predictions.
• The findings may enable financial institutions to better assess risks; however, results are based on a specific dataset and need further validation.