基于电商平台的恶意支付账户识别算法研究-《计算机技术与发展》

文章信息/Info

Title:: Research on Malicious Payment Account Identification Algorithm Based on E-Commerce Platform

Author(s):: FAN Chun-mei; ZHU Jian-sheng; China Academy of Railway Sciences,Beijing 100081,China

Keywords:: E-commerce platform; payment account; k-means; random forest; logistic regression

摘要:: 由于现在电商平台面临着一些虚假交易的现象,为了及时发现在电商平台中经常从事恶意行为的支付账户, 对在电商平台使用的支付账户的交易行为进行了研究,专门针对电商平台可获取的维度进行分析,提出了一种识别电商平台恶意支付账户的方法。该算法基于电商平台销售的特点构建了分类模型的基础特征,使用 k-means 算法进行样本标注,通过关联分析和特征的重要性进行了特征的筛选, 使用随机森林作为分类模型的基础算法。通过对电商平台的真实交易数据进行实验分析,构建的基础特征都具有一定的区分性,使用 k-means 进行聚类的结果,可以很明显地区分恶意支付账户和正常支付账户。对比逻辑回归和随机森林的分类算法,实验结果表明随机森林算法模型具有较高的恶意账户识别准确率和运行效率。

Abstract:: Because of the e-commerce platform facing some false transactions,in order to timely find the payment accounts that often engage in malicious behaviors in the e-commerce platform,some research are carried out on the transaction behaviors of the payment account used in the e-commerce platform. The analysis is performed specially for the available features of the e-commerce platform,and a kind of analysis method of abnormal payment account is proposed. Based on the characteristics of e-commerce platform sales,the basic features of the classification model is built.K-means algorithm is used to mark samples,and features are selected through correlation analysis and the importance of features. Random forest is used as the basic algorithm of the classification model.The experimental analysis by the actual transaction data of the electric business platform can find that the foundation features of built have certain distinction, and the result of the k-means clustering can clearly distinguish between malicious payment account and normal payment account. Compared the classification of the logistic regression algorithm with random forest algorithm model,the experiment shows that random forest algorithm model has higher malicious account recognition accuracy and efficiency.