Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces FCDP, a credit default prediction model that combines an Enhanced Transformer module (ETransformer) for efficient feature filtering and long-range modeling, an Attention Guidance Prediction Module (AGPM) to enhance feature representation and suppress deep feature loss, and a Channel Attention Module (CAM) to learn channel importance. The model addresses limitations in existing credit default prediction research, such as reliance on manual feature engineering and insufficient feature extraction. Experiments on the Lending Club dataset demonstrate that FCDP outperforms six other forecasting models, suggesting its potential for improved risk assessment.
Ditch the feature engineering: a new Transformer-based model predicts credit default with higher accuracy by learning feature importance directly from the data.
Credit default prediction is a key risk control task based on multi-dimensional data such as users' historical behaviour, finance and credit information, and using models to evaluate their future default probability. Aiming at the problems existing in the existing research, such as the performance of default prediction being affected by artificial feature engineering, insufficient feature extraction, and limited ability to capture complex nonlinear dependencies. This paper proposes a credit default prediction model combining an efficient Transformer and channel attention (FCDP). Firstly, the Enhanced Transformer module (ETransformer) is designed to filter redundant feature information, which can improve the computational efficiency of the model while maintaining long-distance modelling and realising high-performance global feature representation. Secondly, the Attention Guidance Prediction Module (AGPM) is designed to enhance the feature representation of the model and effectively suppress the loss of deep features in the data, thus improving the accuracy of large-scale default prediction tasks. Finally, the Channel Attention Module (CAM) is designed so that the model can effectively learn the importance of each channel, combine features more deeply, enhance the extraction of important features and suppress noise data. The experimental results show that FCDP model has been improved on the Lending Club dataset, and it is better than the six forecasting models, which have application potential and can provide more accurate and reliable risk assessment services for credit business.