Open AccessLong short‐term memory based anomaly detection: A case study of China railway passenger ticketing system
Author(s) -
Xie Ze,
Zhu Jiansheng,
Wang Fuzhang,
Li Wen,
Wang Tuo
Publication year - 2021
Publication title -
iet intelligent transport systems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.579
H-Index - 45
eISSN - 1751-9578
pISSN - 1751-956X
DOI - 10.1049/itr2.12007
Subject(s) - anomaly detection , term (time) , anomaly (physics) , computer science , long short term memory , china , automotive engineering , transport engineering , engineering , data mining , artificial intelligence , geography , artificial neural network , physics , archaeology , quantum mechanics , recurrent neural network , condensed matter physics
Data on the network traffic of the railway passenger ticketing system in China are important records that reflect the service status. In this paper, a neural network algorithm based on long short‐term memory using a pre‐wavelet, called the pre‐wavelet long short term memory (PWLSTM), is used that can predict network traffic and forecast its trends over time, with the aim of strengthening anomaly detection in China's railway passenger ticketing service. The proposed algorithm explores the historical regularity and mutation of the data by extracting detailed as well as approximate information on the network traffic. In tests, the results of prediction of the PWLSTM are consistent with long‐ and short‐term regularities in dependencies formed due to the railway passenger ticket‐selling time rule. This verifies that the PWLSTM is suitable for analysing data on China's railway passenger ticketing service. It also records a higher accuracy of prediction than autoregressive models and the LSTM. The PWLSTM is tested by forecasting abnormal network traffic with service anomalies. In the anomaly interval, it is able to detect anomalous points and the prediction results indicate that PWLSTM is a new feasible strategy for passenger ticket service early warning.