Abstract: In recent years, green credit has played an increasingly important role in promoting the construction of China’s green financial system and facilitating high-quality economic development. However, default activities of listed companies in green credit may pose a trust crisis in the financial market and hinder the achievement of the "dual carbon" goals. Therefore, accurately identifying and predicting the green credit default risk of listed companies has become a focus in both academia and the industry. To address this, this paper first constructs a comprehensive measure of default risk, by incorporating the proportion of green credit defaults and the credit ratings of entities. And we process feature engineering by excavating the characteristics of green credit data for listed companies. Subsequently, a hybrid deep learning model, named AE-MACNN-LSTM, is designed by incorporating an AutoEncoder mechanism and various neural networks to predict the green credit default risk of listed companies. Finally, an empirical test is conducted based on samples of A-share listed companies in China. The results indicate that the proposed method achieves significant improvements in multiple evaluation metrics, such as high recall rates and AUC values. This research provides a novel perspective for identifying and mitigating the green credit risk of listed companies, offering valuable insights for the government and investors to timely grasp the dynamics of corporate green credit.

Key words: green credit, default risk predict, hybrid deep learning, AutoEncoder