Explainable Sequential Anomaly Detection via Prototypes (EASD)

EASD: Explainable Sequential Anomaly Detection via Prototypes

We propose EASD, a prototype-based framework for explainable sequential anomaly detection. By linking anomalous subsequences to representative prototypes, EASD provides human-understandable explanations while maintaining strong detection accuracy. Experiments on benchmark log datasets confirm that prototype-based explanations improve interpretability without degrading performance.

Citation

@inproceedings{cheng2023explainable,
  title={Explainable sequential anomaly detection via prototypes},
  author={Cheng, He and Xu, Depeng and Yuan, Shuhan},
  booktitle={2023 International Joint Conference on Neural Networks (IJCNN)},
  pages={1--8},
  year={2023},
  organization={IEEE}
}