Anomaly Access Detection Method Based on Multi-channel Data

Tickets in all ticketing systems become fewer and fewer because more people travel during the holiday. As a result, a group of people who buy tickets on behalf of passengers come into being. The ticket-grabbing behavior of these people has not only put tremendous pressure on ticketing systems, but is profoundly unfair to passengers who want to buy tickets through normal channel. It is necessary to identify malicious accessess of these users in real time and intercept such accesses immediately in order to maintain a better ticketing system. Therefore, an abnormal access structure of the railway ticketing system based on multi-channel data is proposed, which is also helpful to achieve the converged application of rules and models. Through the data service platform, the real-time use of external data and offline data is realized and the accuracy of rule strategy and model is improved. The shared statistical result of rules and model features has reduced the usage of the memory. Moreover, the use of the random forest model can effectively solve problems such as too many rules and poor robustness. It is effective to apply this structure in practical production application.