面向电子商务的用户异常行为检测研究

摘要

In order to detect users’ abnormal trading behavior in e-commerce, users’ behavioral log data were ifrstly classiifed into two categories, namely static attribute sets and operational sequence sets respectively. Then Apriori algorithm based on axis attribute and GSP algorithm for mining sequential patterns were used on these two different data sets, and users’ normal behavior patterns were then established. Finally, user's current behavior patterns and their past normal behavior patterns were compared using pattern matching method based on sequence, and then one could judge whether the user’s trading behavior is normal or not. The experiment on real data sets shows that the method could effectively detect users’ abnormal behavior in e-commerce trading.%针对电子商务中用户异常交易行为的检测问题，文章首先根据用户行为日志数据的特点将其分割为静态属性集和操作序列集，然后利用基于轴属性的Apriori算法和GSP序列模式挖掘算法分别对这两种类型的数据集进行模式挖掘，在此基础上建立用户的正常行为模式，最后使用基于先后顺序的模式比较方法将用户当前的行为模式与其历史正常行为模式进行匹配，以此来判断该用户的交易行为是否异常。在真实数据集上的实验表明，该方法能有效发现电子商务中用户的异常行为。