LogUAD: Log Unsupervised Anomaly Detection Based on Word2Vec

System logs record detailed information about system operation and are important for analyzing the system's operational status performance. Rapid accurate detection of anomalies is great significance to ensure stability. However, large-scale distributed systems becoming more complex, number gradually increases, which brings challenges analyze logs. Some recent studies show that can be unstable due evolution log statements noise introduced by collection parsing. Moreover, deep learning-based methods take a long time train models. Therefore, reduce computational cost avoid instability we propose new Word2Vec-based unsupervised anomaly method (LogUAD). LogUAD does not require parsing step takes original messages as input noise. uses Word2Vec generate word vectors generates weighted sequence feature with TF-IDF handle statements. At last, computationally efficient clustering exploited detect anomaly. We conducted extensive experiments on public dataset from Blue Gene/L (BGL). Experimental results F1-score improved 67.25% compared LogCluster.