Design and Implementation of Mass Log Analysis Method Based on Deep Learning
Yuheng Xing, Yu Zhao, Yulu Zhong · 2020
At present, the Internet of Vehicles platform has become the core infrastructure platform for electric vehicle charging operations, and analyzing the logs of the Internet of Vehicles platform is of great significance for ensuring the stable operation of the charging business and detecting and eliminating system abnormalities. Based on the characteristics of bidirectional long-term and short-term memory neural networks that can effectively use big data and textual context information, this paper proposes a log anomaly detection method. This method first converts the unstructured log into a structured vector, and then inputs it into a two-way long-term and short-term memory network to determine whether an abnormality occurs in the form of two classifications. Experiments show that this method can analyze logs faster and more accurately than traditional methods. Based on this method, operation and maintenance personnel can automate part of their work, and researchers can further explore the automatic isolation and rapid self-healing technology when the Internet of Vehicles platform is abnormal.