Ref:
@inproceedings{Zang2016BeyondST,
title={Beyond Sigmoids: The NetTide Model for Social Network Growth, and Its Applications},
author={Chengxi Zang and Peng Cui and Christos Faloutsos},
booktitle={KDD},
year={2016}
}
More details in the upcoming journal version.
More info on the Social Dynamics and beyond: http://media.cs.tsinghua.edu.cn/~multimedia/cuipeng/
Author: Chengxi Zang
Date: 2017-05-01
C++:
Goal: Generate realistic growth dynamics, both for node and link, of social networks.
The growth dynamics of node and link are captured by NETTIDE equations:Node adoption:
n-dot = beta / t^theta * n * (N - n)
Link build between infected-infected and infected-newly infected:
e-dot = beta_prime / t^theta * n * ( alpha * (n-1)^gamma - e/n ) + 2*n-dot
Assumption:
beta and beta_prime : growth rate of nodes and links; 1/t^theta : fizzling effect the alpha and gamma describe the average number of friends in the ego network.
Two generators:
NetTide-Process: captures the micro-level stochastic interactions within a network.
NetTide-Survival: captures the growth dynamics from hazard rates (hazard process).
Input:
Input network, reads the edge file Data format: blank-separated pairs , e.g 0 1 represents 0->1 Network: e.g.,
1.Complete network, ra ndom network 2.Kronecker network 3.Or any other network
Nodes have been numbered 0-$n-1$ Parameters: beta, theta, beta_prime, alpha, gamma, N
Output:
n(t) over time
e(t) over time. For NetTide-Process, detailed evovling process of nodes and links are recorded.
Matlab:
Goal: Fit the growth dynamics by NetTide model, plot, and regenerate them.