Transforming Temporal-Dynamic Graphs Into Time-Series Data for Solving Event Detection Problems This repository provides a reference implementation for my proposed workflow. This workflow aims to solve event detection problems on temporal-dynamic graphs. You can find comlete datasets (size>100MB) in Link.
Detailed walktrough can be found in Proposed_Workflow.ipynb file.
python>=3.8
networkx
numpy
pandas
gensim==3.8.3
node2vec
matplotlib
holoviews
sklearn
scipy
merlion (InstallGuide is bellow)
This implementation uses Merlion library for unsupervised time-series anomaly detection alogrithms. You can insatall the necessary libraries with using pip install salesforce-merlion[all]. It is important to use [all] to able to use deep learning based methods.
For further information:
https://github.com/salesforce/Merlion
In this study we have used tdGraphEmbed algorithm. Code in this repository is directly using the tdGraphEmbed implementation.
Futher information and source codes are available at:
https://github.com/moranbel/tdGraphEmbed
Paper - Link