zh-shuai / multiad

Implementation of "Multiple Hypothesis Testing for Anomaly Detection in Multi-type Event Sequences" (ICDM 2023)

MultiAD: Multiple Hypothesis Testing for Anomaly Detection in Multi-type Event Sequences

The implementation of our ICDM-2023 paper "Multiple Hypothesis Testing for Anomaly Detection in Multi-type Event Sequences".

Installation

1. Install the dependencies

   conda env create -f environment.yml
   

2. Activate the conda environment

   conda activate anomaly_mpp
   

3. Install the package (this command must be run in the MultiAD folder)

   pip install -e .
   

4. Unzip the data

   unzip data.zip
   

Reproducing the results from the paper

• experiments/spp.py: GOF testing for the standard Poisson process (Section V-A in the paper).
• experiments/multivariate.py: Detecting anomalies in synthetic data (Section V-B).
• experiments/real_world.py: Detecting anomalies in real-world data (Section V-C).

Citation

If you find this code useful, please consider citing our paper. Thanks!

@inproceedings{zhang2023multiple,
  title={Multiple Hypothesis Testing for Anomaly Detection in Multi-type Event Sequences},
  author={Zhang, Shuai and Zhou, Chuan and Zhang, Peng and Liu, Yang and Li, Zhao and Chen, Hongyang},
  booktitle={2023 IEEE International Conference on Data Mining (ICDM)},
  pages={808--817},
  year={2023},
  organization={IEEE}
}

Acknowledgements and References

Parts of this code are based on and/or copied from the code of: https://github.com/shchur/tpp-anomaly-detection, of the paper "Detecting Anomalous Event Sequences with Temporal Point Processes".