A co-evolving multilayer model of the flow of the news.
For further details about the model, its implementation, advantages and limitations, and empirical results, please refer to the original paper [1].
The packages python-igraph (version 0.7.1) and cyrandom (version 0.3.1) must be installed in order to run the NewsFlow, which can be easily done with the pip command:
$ pip install python-igraph
$ pip install cyrandom
(for troubleshooting, refer to the python-igraph user manual).
First, clone the git repository:
$ git clone https://github.com/ikicab/NewsFlow/
Then navigate to the downloaded folder and type the following commands into the terminal:
$ python setup.py build_ext --inplace
$ cython -a NewsFlow.pyx
The algorithm accepts the following input parameters – note that only the input graph g, the initial number F0 of consumers of false news, the news-alignment rate eta, and the spreading rate of false news lambda_F are mandatory.
| Parameter name | Variable type | Description | Default value (if applicable) |
|---|---|---|---|
g |
igraph.Graph |
population layer in the form of an (unweighted undirected) graph | |
F0 |
int |
initial number of consumers of false news | |
eta |
double |
rate of news-alignment (news-rewiring) | |
lambda_F |
double |
spreading rate of false news | |
lambda_T |
double |
spreading rate of true news | 1 |
M |
int |
number of media outlets that disseminate false/true news content; in total, there are 2*M news providers |
10 |
t_max |
int |
upper limit on the simulation time (ignored if 0) |
0 |
window_size |
int |
length of the sliding window over which the sample variance in the number of unbalanced (active) edges/polarised news consumers is computed (ignored if 0) |
0 or len(g.vs) if t_max == 0 |
tol |
double |
tolerance on the sliding-window variance (ignored if window_size == 0) |
0.02 |
sfreq |
int |
state-saving frequency | 0 |
Output parameters depend on the choice of sfreq; see below.
Applicable if sfreq |
Parameter name | Variable type | Description |
|---|---|---|---|
== 0 |
time_range |
list |
time-step evolution |
== 0 |
list_polarised |
list |
detailed time evolution of the number of polarised individuals |
== 0 |
list_rhoF |
list |
detailed time evolution of the number of consumers of false news |
== 0 |
list_FT_edges |
list |
detailed time evolution of the number of unbalanced (active) edges |
< 0 |
v_polarised |
int |
final number of polarised individuals |
< 0 |
rhoF |
int |
final number of consumers of false news in the population |
< 0 |
nFT_edges |
int |
final number of unbalanced (active) edges |
> 0 |
list_rhoF |
list |
detailed time evolution (with a sampling frequency of sfreq) of the number of consumers of false news |
> 0 |
X_Ymedia |
dict |
detailed time evolution (with a sampling frequency of sfreq) of the number of media outlets that disseminate Y news content subscribed to by consumers of X news (with X and Y standing for true/false) |
> 0 |
Xmedia_Y |
dict |
detailed time evolution (with a sampling frequency of sfreq) of the number of consumers of Y news subscribed to media outlets that disseminate X news content (with X and Y standing for true/false) |
A selection of dedicated visualisation tools may be found in the accompanying Python script VisualTools (refer also to the Tutorial).
See the Tutorial for more help and examples.
[1] B. Ikica, Co-evolution of the Multilayer News Flow. In preparation; a detailed overview can be found here (refer to pages 161–199).
The figure above was prepared with the LaTeX drawing package tikz-network.