Research on Artificial Intelligence Driven Anomaly Detection Model for Big Data

Abstract

Facing the high-dimensionality, heterogeneity and temporal complexity of anomaly detection in big data environment, an intelligent detection model integrating graph neural network, self-encoder and attention mechanism is designed. The model structure is equipped with multimodal feature encoding capability and online adaptive mechanism, which improves the recognition performance of rare anomalies and structural mutations. Experiments based on the KDDCup99 and NSL-KDD datasets demonstrate that the model outperforms multiple comparative methods in terms of accuracy and robustness, and shows good practicality and scalability.