Tie-decay temporal networks

Code accompanying "Tie-decay temporal networks in continuous time and eigenvector-based centralities" by Walid Ahmad, Mason Porter, and Mariano Beguerisse-Díaz. ¹

This repository contains utilities for loading and computing tie-decay centrality scores for temporal networks.

Setup

conda env create -f conda_environment.yml
python setup.py develop

Load Dataset

from tiedecay.dataset import Dataset

raw_data = [(1, 5, "2020-01-01-00:01:23"), (3, 2, "2019-08-12-11:01:34"), ...]
user_mapping = {1: "henry ford",  2: "nikola tesla", ...}

dataset = Dataset(raw_data, user_mapping)

Compute centrality scores

from tiedecay.construct import TieDecayNetwork

# half-life of one day
alpha = np.log(2)/24/3600
tdn = TieDecayNetwork(dataset, alpha=alpha)

Compute centrality values at sampled time points

centrality_df = tdn.compute_centrality_trajectories_from_dataset(100, 'pagerank')

Compute the tie-decay matrix at a given time

t_select = "2020-01-02-12:00:00"
B_t = tdn.compute_from_dataset(t_select)

References

[1] arXiv preprint, 2018 arXiv:1805.00193v2