Psi-score

$\psi$-score: Metric of user influence in Online Social Networks

Requirements

• Python >=3.9,<3.11

Installation

$ pip install psi-score

Usage

>>> from psi_score import PsiScore
>>> adjacency = {0: [1, 3], 1: [0, 2], 2: [0, 1, 3], 3: [0]}
>>> lambdas = [0.23, 0.50, 0.86, 0.19]
>>> mus = [0.42, 0.17, 0.10, 0.37]
>>> psiscore = PsiScore()
>>> scores = psiscore.fit_transform(adjacency, lambdas, mus)
>>> scores
array([0.21158803, 0.35253745, 0.28798439, 0.14789014])
>>> np.round(scores, 2)
array([0.21, 0.35, 0.29, 0.15])

You can use another algorithm and change some parameters:

>>> psiscore = PsiScore(solver='power_nf', n_iter=500, tol=1e-3)
>>> scores = psiscore.fit_transform(adjacency, lambdas, mus, ps=[1], qs=[0, 3])

The ps and qs parameters allows to have some chosen p_i and q_i vectors (only with the push and power_nf methods):

>>> psiscore.P
{1: array([0.5333334 , 0.1681094 , 0.46801851, 0.34442264])}
>>> psiscore.Q
{0: array([0.46164044, 0.0514935 , 0.02798624, 0.30484491]),
 3: array([0.13087053, 0.01616898, 0.01850541, 0.42554885])}

License

psi-score was created by Nouamane Arhachoui. It is licensed under the terms of the MIT license.

References

• Giovanidis, A., Baynat, B., Magnien, C., & Vendeville, A. (2021). Ranking Online Social Users by Their Influence. IEEE/ACM Transactions on Networking, 29(5), 2198-2214. https://doi.org/10.1109/tnet.2021.3085201

• Arhachoui, N., Bautista, E., Danisch, M., & Giovanidis, A. (2022). A Fast Algorithm for Ranking Users by their Influence in Online Social Platforms. 2022 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), 526-533. https://doi.org/10.1109/ASONAM55673.2022.10068673