Cross-border e-commerce credit risk early warning model based on deep learning

Credit risk challenges in cross-border e-commerce are becoming more pronounced, significantly hindering the sector's healthy progression. Existing early warning mechanisms cannot effectively analyse the complex, nonlinear patterns and dynamic changes in cross-border e-commerce transactions, causing low detection accuracy. To this end, this study first introduces an enhanced SMOTE algorithm to address class imbalance in credit hazard data, followed by the selection and standardisation of key hazard indicators. Then one-dimensional CNN is used to deeply mine the features of impact indicators. The particle swarm optimisation (PSO) algorithm is utilised to fine-tune the parameters of the extreme learning machine (ELM), with the optimised ELM subsequently generating the risk warning classifications. The experimental outcome indicates that the forecasting accuracy of the offered model is improved by at least 4.05%, demonstrating significant gains in warning system accuracy.