深度学习模型在跨境支付信息体系中的应用探究——基于自由格式报文分类场景  

Exploration of the Application of Deep Learning Models in Cross-border Payment and Clearing Business:Based on the Classification Scenario of Inquiry and Reply Messages

作　　者：交通银行金融服务中心支付清算课题组 郑英姿 Research Group of Payment and Clearing of Financial Service Center Dept.of BOCOM

机构地区：[1]不详 [2]交通银行金融服务中心

出　　处：《新金融》2025年第3期48-57,共10页New Finance

摘　　要：跨境支付信息系统是银行间结算信息传递的重要基础。随着跨境贸易增长催生的支付需求增长,人工处理业务的效率和成本问题逐渐凸显。深度学习技术在学习文本信息的语义表示和特征提取上优势明显,是文本分类常用的技术方法。本文分别训练了RNN、LSTM、BERT三个深度学习模型对跨境自由格式报文进行分类,并对模型之间的性能差异进行了比较。试验结果表明,深度学习技术在基于跨境汇款的支付自由格式报文分类任务中的准确率较高,报文处理速度大幅提升。商业银行可在控制风险的前提下应用深度学习技术,利用预训练模型对自由格式报文进行智能处理,推动数字化转型,降低运营成本。The cross-border payment information system is an essential foundation for the transmission of inter-bank settlement information.The growth in cross-border trade leads to a rise in demand for cross-border payments,and the efficiency and cost issues of manual processing are gradually highlighted.Deep learning technology,which is often used as a common method for text classification,has obvious advantages in learning semantic representations and extracting features from text information.Through the training of three models-RNN,LSTM,and BERT,this article constructed a comprehensive comparison of the performance of the models,based on the classification of the cross-border free format messages.The experimental results found that the deep learning technology achieves high accuracy,and significantly improves the processing speed of the cross-border free format messages.Commercial banks,on the premise of controlling risks,could apply the deep learning technology to intelligently process the free-format messages.By using pre-trained models,commercial banks could promote the digital industry upgrading,and reduce the operational costs.

关 键 词：深度学习 金融科技 自然语言处理 跨境支付 

分 类 号：F830[经济管理—金融学]