结合遗传算法和集成学习的信用卡财务欺诈交易检测  

Detecting Credit Card Financial Fraudulent Transactions Using Genetic Algorithm and Ensemble Learning

作　　者：薛明香 XUE Ming-xiang(School of Big Data and Economic Management,Hefei Professional College of Economics and Technology,Anhui,Hefei 230031)

机构地区：[1]合肥经济技术职业学院大数据与经济管理学院,安徽合肥230031

出　　处：《贵阳学院学报(自然科学版)》2025年第1期81-86,91,共7页Journal of Guiyang University:Natural Sciences

基　　金：2020年安徽省高校人文社科研究重点项目(SK2020A0939);2021年安徽省高校人文社会科学研究项目(SK2021A1171)。

摘　　要：随着数字化和信息化技术的发展,线上金融交易已被广泛应用。随之而来的欺诈交易也为金融机构和企业的财产安全带来了巨大威胁,迫切需要有效的检测方法,特别是检测信用卡欺诈对于识别和防止未经授权的交易至关重要。为此,提出了集合遗传算法和学习的欺诈交易检测方法。首先,通过欠采样和合成少数过采样(SMOTE)技术,解决信用卡数据集的数据不平衡问题。其次,所提方法智能地结合了多种算法,包括随机森林(RF)、K最近邻(KNN)和多层感知器(MLP)分类器,并通过遗传算法(GA)进行适当的加权优化,以增强欺诈识别能力。在公开信用卡交易数据集上的实验结果表明,所提集成模型在精度、召回率和F1得分等指标上均取得了比已有机器学习方法和单个分类器更好的性能,证明了集成学习方法在欺诈交易检测中的有效性。With the development of digital and information technologies,online financial transactions have been widely adopted.Along with this,fraudulent transactions pose a significant threat to the financial security of institutions and enterprises,necessitating effective detection methods.Detecting credit card fraud is crucial for identifying and preventing unauthorized transactions.Therefore,this paper proposes a method for detecting fraudulent transactions by combining genetic algorithm and ensemble learning.Firstly,the method addresses the data imbalance issue of credit card datasets through under-sampling and Synthetic Minority Over-sampling Technique(SMOTE).Secondly,the proposed approach intelligently integrates multiple algorithms,including Random Forest(RF),K-Nearest Neighbors(KNN),and Multilayer Perceptron(MLP)classifiers,and optimizes the weights using genetic algorithm to enhance fraud detection capability.Experimental results on a publicly available credit card transaction dataset demonstrate that the proposed ensemble model achieves better performance in terms of precision,recall,and F1-score compared to existing machine learning methods and individual classifiers,validating the effectiveness of ensemble learning methods in fraudulent transaction detection.

关 键 词：欺诈交易检测 遗传算法 集成学习 合成过采样 支持向量机 K最近邻 多层感知器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]