TY - GEN
T1 - An Evaluation of Log Anomaly Detection with Log Event Embedding
AU - Zhang, Yucheng
AU - Li, Cheng
AU - Chen, Qian
AU - Zhou, Hongwei
AU - Wang, Chao
AU - Tian, Bowen
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - To enhance the accuracy of log anomaly detection, the raw log is transformed into word vectors with the different word embedding algorithms. Currently, some kinds of word embedding algorithms such as GloVe have been applied in log anomaly detection. However, there is no academic consensus on which word embedding algorithm is the most suitable for log anomaly detection. Addressing this issue, we have developed an experimental platform for log anomaly detection, aiming to comparatively analyze the different word embedding algorithms in log anomaly detection. We first preprocessed the Loghub dataset, then generated log vectors using GloVe, Word2vec, and FastText, respectively. Finally, we trained an LSTM model for anomaly detection. Through in-depth analysis of the experimental data, this paper evaluates the different word embedding algorithms in log anomaly detection. The experimental results indicate that, GloVe achieves the highest accuracy under the environment in this paper.
AB - To enhance the accuracy of log anomaly detection, the raw log is transformed into word vectors with the different word embedding algorithms. Currently, some kinds of word embedding algorithms such as GloVe have been applied in log anomaly detection. However, there is no academic consensus on which word embedding algorithm is the most suitable for log anomaly detection. Addressing this issue, we have developed an experimental platform for log anomaly detection, aiming to comparatively analyze the different word embedding algorithms in log anomaly detection. We first preprocessed the Loghub dataset, then generated log vectors using GloVe, Word2vec, and FastText, respectively. Finally, we trained an LSTM model for anomaly detection. Through in-depth analysis of the experimental data, this paper evaluates the different word embedding algorithms in log anomaly detection. The experimental results indicate that, GloVe achieves the highest accuracy under the environment in this paper.
KW - LSTM
KW - log anomaly detection
KW - word embedding
UR - https://www.scopus.com/pages/publications/105021823771
U2 - 10.1109/ICNC-FSKD67701.2025.11198098
DO - 10.1109/ICNC-FSKD67701.2025.11198098
M3 - 会议稿件
AN - SCOPUS:105021823771
T3 - ICNC-FSKD 2025 - 21st International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery
SP - 577
EP - 582
BT - ICNC-FSKD 2025 - 21st International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery
A2 - Zhao, Liang
A2 - Xiao, Zheng
A2 - Li, Kenli
A2 - Wang, Lipo
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 21st International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery, ICNC-FSKD 2025
Y2 - 26 July 2025 through 28 July 2025
ER -