Anomaly Detection of E-commerce Econnoisseur Based on User Behavior

Abstract

Econnoisseur refers to users who obtain high returns from the Internet at low cost. It is of great significance for platform to identify econnoisseur to reduce unnecessary losses. At present, econnoisseur is mainly intercepted by rules. This method will fail when the new get the best deal method appears, and there is a certain lag. This paper identifies the econnoisseur from Knownsec Security Intelligence Brain’s e-commerce website visitors. First of all, it is found that the precision and recall of the Isolation Forest are better than the Local Outlier Factor and DBSCAN in econnoisseur detection. Secondly, we merged the similar URLs visited by users with Bi-directional Long Short-Term Memory (BiLSTM), then use the merged data in Isolation Forest Model. It is found that the improved Isolation Forest model based on BiLSTM can further improve the detection ability. Practical case studies showed that this method has certain validity and reference for the detection of econnoisseur.