sabcom: A Jupyter Notebook repository from mapleyustat

The Spatial Agent-Based Covid Model (SABCom)

SABCoM is an open source easy-to-use-and-adapt spatial network multi-agent simulation model of the spread of the Covid-19 virus. The model can be used to simulate and analyse a spread of the Covid-19 virus on the level of the individual. This distinguishes the model from most models, that have a more high level perspective. This makes our model particularly usefull for researchers and policy makers that want to study the impact of heterogeneity in the population (e.g. by modelling neighbourhoods with vulnerable populations) and who want to study the impact of targeted measures (e.g. quarantining certain neighbourhoods, or targetted social distancing measures at aimed at specific segments of the population).

The model is inspired by the canonical SEIRS structure and generates curves that reflect the amount of agents that are susceptible (s) infected without symptoms (i1), with symptoms (i2), critically ill (c), and are recovered (r).

Due to the unique spatial structure of the model, we can track how a virus spreads spatially. For example through Cape Town. The figure below shows the quantity of the population infected in different Wards in the City of Cape town. Note that thisconcerns a hypothetical simulation of the non-calibrated model and is only used to give an idea of possible dynamics.

Getting started

You can install the SABCom model by cloning this repository to your system. After that, you are ready to start using the model.

Running the model

There are two files in the main folder that will allow you to run the model. In the first step, you set the parameters for the simulation and store them in a json file. Open SABCoModel_notebook.ipynb and set

parameters = {
    # general simulation parameters
    "time": 150,
    "number_of_agents": 10000,
    ...
}

and save as json file with

with open('parameters.json', 'w') as outfile:
    json.dump(parameters, outfile)

Next, you also need to set the location and number of neigborhoods. We use population density data from 116 Wards in and around Cape Town.

population = pd.read_csv('population.csv')
smallest_size = population['Population'].sum() / parameters['number_of_agents']
neighbourhood_data = []
for i in range(len(population)):
    if population['Population'].iloc[i] > smallest_size:
        neighbourhood_data.append(
            [int(population['WardID'].iloc[i]), {'Population': population['Population'].iloc[i],
                                                            'Density': population['Density'].iloc[i],
                                                            'lon': population['lon'].iloc[i],
                                                            'lat': population['lat'].iloc[i]}])
max_neighbourhoods = len(neighbourhood_data)
with open('neighbourhood_data.json', 'w') as outfile:
    json.dump(neighbourhood_data[:max_neighbourhoods], outfile)

The configuration files parameters.json and neighbourhood_data.json can then be fed into the main program. There are two options:

you run the Initialization, Simulation and Save network data cells from within the notebook
you run python SABCoModel.py

Analysis The notebook contains a section that produces the output graphs

Requirements The program requires python 3 and the geopy, networkx non-standard packages.

Disclaimer

This software is intended for educational and research purposes. Despite best efforts, we cannot fully rule out the possibility of errors and bugs. The use of SABCoM is entirely at your own risk.

mapleyustat/sabcom