An Ensemble Model Minimising Misjudgment Cost: Empirical Evidence From Chinese Listed Companies

Abstract

Predicting corporate financial distress is critical for bank lending and corporate bond investment decisions. Incorrect identification of default status can mislead lenders and investors, leading to substantial losses. This paper proposes an ensemble model that minimises the overall cost of misjudgment by considering the imbalanced ratio weighted loss of the unbalanced ratio of Type I and Type II errors in the objective function. Unlike existing static financial distress prediction models, the proposed model integrates panel data by using time-shifting to account for credit risk dynamics. To validate the prediction model, data were collected for Chinese listed companies, considering geographic area, ownership structure and firm size. We demonstrate that by weighting predictions from different classification models, the overall misjudgment cost can be minimised. This study identifies earnings per share and the product price index as the most relevant indicators affecting the financial performance of Chinese-listed companies. Overall, the results indicate that the proposed model has a predictive capacity of up to 5 years, with 98.7% for 1-year forecasting horizons and 96.8% for 5-year-ahead forecasting horizons. Furthermore, the proposed model outperforms existing distress prediction models in overall prediction performance by correctly identifying defaulting companies while avoiding misjudging good companies.