基于集成学习的金融交易欺诈识别研究  

Research on Financial Transaction Fraud Identification Based on Ensemble Learning

作　　者：郑德铭 李思佳 潘彦恺 郑健龙 ZHENG Deming;LI Sijia;PAN Yankai;ZHENG Jianlong(Graduate School,China People's Police University,Langfang 065000,China;Smart Policing College,China People's Police University,Langfang 065000,China)

机构地区：[1]中国人民警察大学研究生院,河北廊坊065000 [2]中国人民警察大学智慧警务学院,河北廊坊065000

出　　处：《现代信息科技》2025年第4期173-178,共6页Modern Information Technology

基　　金：河北省社会科学基金项目(HB22SH011)。

摘　　要：金融欺诈严重威胁金融市场稳定,而现有的反欺诈手段存在单一性和低效率的问题。为此,文章基于集成学习方法构建了金融交易欺诈识别模型,旨在提升欺诈识别效果。研究中采用装袋法(Bagging)和提升法(Boosting)构建了4个基础模型,并通过优化参数筛选出2个效果较好的模型。随后,利用堆叠法(Stacking)对这2个模型进行融合训练,进一步提高了模型的识别率。实验结果表明,融合模型在金融交易欺诈识别中具有显著优势。与基础模型相比,其在不同数据集上的准确率更高,尤其在处理复杂欺诈模式和新型手段时,展现出更高的准确性和稳定性。这种改进的模型方法为金融决策者和相关部门提供了有效的决策支持,有助于提升金融市场的安全性。Financial fraud seriously threatens the stability of financial markets,and the existing anti-fraud methods have the problems of singleness and inefficiency.Therefore,this paper constructs a financial transaction fraud recognition model based on the Ensemble Learning method,aiming to improve the fraud recognition effect.In the research,four basic models are constructed by Bagging and Boosting,and two models with better effects are selected by optimizing parameters.Subsequently,the Stacking method is used to conduct fusion training for the two models,which further improves the recognition rate of the model.The experimental results show that the fusion model has significant advantages in financial transaction fraud identification.Compared with the basic model,it has higher accuracy with different datasets,especially in dealing with complex fraud patterns and new means,showing higher accuracy and stability.This improved model method provides effective decision support for financial decision makers and relevant departments,and helps to improve the security of financial markets.

关 键 词：集成学习 金融欺诈 BOOSTING STACKING 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] F830[自动化与计算机技术—控制科学与工程]