zjlxgxz / DynAnom

Codebase for KDD22 paper "Subset Node Anomaly Tracking over Large Dynamic Graphs"

Codebase for Subset Node Anomaly Tracking over Large Dynamic Graphs @KDD-2022

Authors: Xingzhi Guo, Baojian Zhou and Steven Skiena

0. Download data:

Download the compiled Person-Event Graph, and unzip them into the folder toy-data

• Person-Event Graph Link:

 https://drive.google.com/drive/folders/1nu-2Lx80WGD9I7cjXZ1H1SKauTr_UIe7?usp=sharing

1. Set up env

sh ./recipe/setup.sh

Notes 1. Due to weiredness of Numba, a specific python version (3.7) is needed: #SEE: numba/numba#5156

Notes 2. That is a weighted version of DynamicPPE

2. Run the method (all experiments)

sh ./recipe/run.sh

3. Output figure/results:

Once get all results in ./output Run all cells in:

./src/visualize.ipynb

The figures/results should be reproduced. Cheers!

4. Please kindly support us by citing the following papers:

@inproceedings{guo2022dynanom,
  title={Subset Node Anomaly Tracking over Large Dynamic Graphs},
  author={Guo, Xingzhi and Zhou, Baojian and Skiena, Steven},
  year={2022}, 
  booktitle = {Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery & Data Mining}, 
  series = {KDD '22},
  doi = {10.1145/3534678.3539389}, 
  publisher = {Association for Computing Machinery}, 
  url = {https://doi.org/10.1145/3534678.3539389}, 
}

@inproceedings{guo2021subset,
  title={Subset Node Representation Learning over Large Dynamic Graphs},
  author={Guo, Xingzhi and Zhou, Baojian and Skiena, Steven},
  year={2021}, 
  booktitle = {Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining}, 
  pages = {516–526}, 
  numpages = {11}, 
  series = {KDD '21},
  doi = {10.1145/3447548.3467393}, 
  publisher = {Association for Computing Machinery}, 
  url = {https://doi.org/10.1145/3447548.3467393}, 
}

@inproceedings{zhang2016approximate,
  title={Approximate personalized pagerank on dynamic graphs},
  author={Zhang, Hongyang and Lofgren, Peter and Goel, Ashish},
  booktitle={Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining},
  pages={1315--1324},
  year={2016}
}

@inproceedings{andersen2006local,
  title={Local graph partitioning using pagerank vectors},
  author={Andersen, Reid and Chung, Fan and Lang, Kevin},
  booktitle={2006 47th Annual IEEE Symposium on Foundations of Computer Science (FOCS'06)},
  pages={475--486},
  year={2006},
  organization={IEEE}
}