Research on log anomaly detection method based on deep learning

In this paper, we construct a deep learning detection framework that integrates multi-source information, temporal modeling and semantic comparison mechanism, covering four modules: log preprocessing, representation learning, model construction and optimization. In the model layer design, BiGRU and attention mechanism are introduced to capture global contextual dependencies, contrast learning and twin networks are combined to enhance feature discrimination, and regular optimization and policy scheduling are used to improve model stability. Empirical studies on real industrial log datasets, such as HDFS, BGL and OpenStack, show that this paper’s model significantly outperforms the existing mainstream methods in terms of accuracy, F1-score and AUC, and has good generalization ability and anomaly detection performance.