Skip to content

TGX logo

Temporal Graph Analysis with TGX (WSDM 2024 Demo Track)

TGX supports all datasets from TGB and Poursafaei et al. 2022 as well as any custom dataset in .csv format. TGX provides numerous temporal graph visualization plots and statistics out of the box.

Data Loading

For detailed tutorial on how to load the datasets into tgx.Graph, see docs/tutorials/data_loader.ipynb

  1. Load TGB datasets
import tgx
dataset = tgx.tgb_data("tgbl-wiki")
ctdg = tgx.Graph(dataset)
  1. Load built-in datasets
dataset = tgx.builtin.uci()
ctdg = tgx.Graph(dataset)
  1. Load custom datasets from .csv
from import read_csv
toy_fname = "docs/tutorials/toy_data.csv"
edgelist = read_csv(toy_fname, header=True,index=False, t_col=0,)

Visualization and Statistics

For detailed tutorial on how to generate visualizations and compute statistics for temporal graphs, see docs/tutorials/data_viz_stats.ipynb

  1. Discretize the network (required for viz)
dataset = tgx.builtin.uci()
ctdg = tgx.Graph(dataset)
time_scale = "weekly"
dtdg, ts_list = ctdg.discretize(time_scale=time_scale, store_unix=True)
  1. Plot the number of nodes over time
tgx.degree_over_time(dtdg, network_name="uci")
  1. Compute novelty index

Install dependency

Our implementation works with python >= 3.9 and can be installed as follows

  1. set up virtual environment (conda should work as well)
python -m venv ~/tgx_env/
source ~/tgx_env/bin/activate
  1. install external packages
pip install -r requirements.txt
  1. install local dependencies under root directory /TGX
pip install -e .
  1. [alternatively] install from test-pypi
pip install -i py-tgx

You can specify the version with ==, note that the pypi version might not always be the most updated version

  1. [optional] install mkdocs dependencies to serve the documentation locally
pip install mkdocs-glightbox

Creating new branch

first create the branch on github

git fetch origin

git checkout -b test origin/test