[1]郝晓培,朱建生,单杏花.图神经网络在 12306 黑产用户挖掘的研究[J].计算机技术与发展,2022,32(07):185-190.[doi:10. 3969 / j. issn. 1673-629X. 2022. 07. 032]
　HAO Xiao-pei,ZHU Jian-sheng,SHAN Xing-hua.Research on Graph Neural Network in 12306 Black Production User Mining[J].,2022,32(07):185-190.[doi:10. 3969 / j. issn. 1673-629X. 2022. 07. 032]
点击复制

图神经网络在 12306 黑产用户挖掘的研究()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
32
期数:
2022年07期
页码:
185-190
栏目:
应用前沿与综合
出版日期:
2022-07-10

文章信息/Info

Title:
Research on Graph Neural Network in 12306 Black Production User Mining
文章编号:
1673-629X(2022)07-0185-06
作者:
郝晓培; 朱建生; 单杏花
中国铁道科学研究院,北京 100081
Author(s):
HAO Xiao-pei; ZHU Jian-sheng; SHAN Xing-hua
China Academy of Railway Sciences,Beijing 100081,China
关键词:
黑产用户; 个体特征; 社交网络; 图神经网络; 邻居节点; 支持向量机
Keywords:
black production users; individual characteristics; social networks; graph neural network; neighbor node; support vector machine
分类号:
TP301. 6
DOI:
10. 3969 / j. issn. 1673-629X. 2022. 07. 032
摘要:
随着铁路信息化技术的高速发展以及铁路互联网售票系统的不断优化完善,12306 已成为铁路客运主要的售票渠道,为旅客的出行带来了极大的便利,然而节假日部分线路供需仍存在巨大缺口,铁路客票销售市场存在巨大的牟利空间,从而也面临着网络黑色产业链的威胁。 针对当前铁路 12306 互联网售票系统存在黑产用户抢票,倒票,囤票等问题,提出了兼顾旅客社会关系以及个体特征的黑产用户识别模型。 首先基于旅客的历史购票及出行行为,从时间、空间等维度构建旅客个体特征,然后基于旅客的出行关系以及购票关系构建旅客社交网络,通过频率反映旅客社交关系强度,最后,采用图神经网络将节点个体特征以及邻居节点的特征信息线性表示为低维稠密的向量空间,将其最终旅客特征向量输入无核二次曲面支持向量机进行黑产用户识别。 实验表明,综合考虑旅客社交关系以及旅客个体特征的黑产用户识别模型相对于只考虑个体特征的模型准确率有了显著的提高。
Abstract:
With the rapid development of railway information technology and the continuous optimization and improvement of the railway Internet ticketing system,12306 has become the main ticketing channel for railway passenger transportation, bringing great convenience to passengers. However,there is still a huge gap in supply and demand on some routes? during holidays. There is a huge profit - making space in the railway ticket sales market,which also faces the threat of a black network industry chain. Regarding the current railway12306 Internet ticketing system,there are problems such as black-produced users grabbing tickets,scalping tickets,and hoarding tickets.We propose a black product user identification model that takes into account the social relationship of passengers and individual characteristics. Firstly,based on the historical ticket purchase and travel behavior of passengers,the individual characteristics of passengers are constructed from the dimensions of time and space,and then the passenger social network is constructed based on the travel relationship and ticket purchase relationship of the passengers, and the intensity of the social relationship of the passengers is reflected by frequency.Finally,graph nerves are used. The network linearly expresses the individual characteristics of nodes and the characteristic information of neighboring nodes as a low-dimensional dense vector space,and inputs the final passenger feature vector into the coreless quadric support vector machine for black user identification. Experiments show that the black product user identification model that comprehensively considers the? social relationship of passengers and the individual characteristics of passengers has a significant improvement in accuracy compared with the model that only considers individual characteristics.
更新日期/Last Update: 2022-07-10