摘要
为了解决电子商务平台中存在的虚假交易问题,本文依据商品的销售记录以及商家的基本信息,提出了一种结合深度置信网络和多层感知器的虚假交易识别方法,通过识别出以通过刷单增加销量的商品来识别虚假交易。首先利用深度置信网络对交易特征进行学习,得到更高层次的抽象特征;然后利用多层感知器进行分类任务,从而识别出虚假交易。从淘宝中爬取商品的交易记录和评论数据进行实验验证,与其他机器学习模型的实验结果进行对比,其性能有明显的提升。
For solving the problem of fraud transaction in e-commerce platform,a method that combined Deep Belief Networks and Multilayer Perceptron based on the transaction records and review records of Products was put forward.Through recognizing the product which was increased sales in fraudulent transactions to recognize the fraud transactions.The features of transaction were learned by DBN to get the higher level of abstract features,and the MLP performed the classification task.Tested by experiments using the transaction records and review records of products crawled from Taobao,the comprehensive performance had improved significantly compared with the other machine learning model.
作者
刘畅
殷聪
Liu Chang Yin Cong(School of Information Management, Wuhan University, Wuhan 430072, Chin)
出处
《现代情报》
CSSCI
北大核心
2016年第10期62-67,73,共7页
Journal of Modern Information