This repository accompanies the article "On limitations of uniplex networks for modeling multiplex contagion" by Nicholas W. Landry and jimi adams and provides all scripts necessary to reproduce all results and figures.
This repository provides the bases for a project seeking to demonstrate the limitations of modeling multiplex diffusion processes from uniplex network data. We do so by: starting from multiplex data sources, and simulating diffusion processes over those networks. We then decompose each network into uniplex versions, and re-simulate the same diffusion processes over those. We then examine how the addition of the component uniplex simulations correspond to the composite simulation process.
What's here:
Data (in the Data folder)
-
The import_CO90_data_full.py and import_CO90_data_GC.py file convert the raw data in the ICPSR_22140 folder from "Project 90" in Colorado Springs (original data) into NetworkX objects and saves using the Python "shelve" module. They load the full data and the giant component respectively. The output objects can be retrieved with the following keys:
- sex-network - NetworkX object of the sex network
- drug-network - NetworkX object of the drug network (both drugs and needles)
- combined-network - NetworkX object of the multiplexed sex and drug networks
-
The import_JOAPP_data_full.py and import_JOAPP_data_GC.py files convert the raw data in the Twitter-Foursquare folder (original data) into NetworkX objects and saves using the Python "shelve" module. They load the full data and the giant component respectively. The output objects can be retrieved with the following keys:
- foursquare-network - NetworkX object of the Foursquare network
- twitter-network - NetworkX object of the Twitter network
- combined-network - NetworkX object of the multiplexed Foursquare and Twitter networks
Libraries
- multiplex_contagion.py is a collection of functions to simulate contagion processes on multiplexed and uniplex data
- plot_utilities.py contains the function used to plot the multiplex networks in both Jupyter notebooks.
Scripts
- run_SI_time_series.py runs an ensemble of SI contagion simulations for a list of p values and outputs the time series for each realization and parameter value into a shelve file.
- run_threshold_time_series.py runs an ensemble of threshold contagion simulations for a list of threshold values and outputs the time series for each realization and parameter value into a shelve file.
- plot_SI_time_series.py plots the result from run_SI_time_series.py
- plot_threshold_time_series.py plots the result from run_threshold_time_series.py
Notebooks
- plot_CO90_dataset.ipynb generates a multiplex visualization of the Project 90 dataset based on this script
- plot_JOAPP_dataset.ipynb generates a multiplex visualization of the JOAPP dataset based on this script
Getting Started
Clone this repository and within the main directory, make a directory titled "Data". Download the Project 90 dataset and the JOAPP dataset. Within the Data folder, copy the unzipped ICPSR_22140 and rsos160863_si_001 folders from the Project 90 and JOAPP datasets respectively into the Data folder. All scripts should work once this is complete.