popm: Public Order Policing Model

Developer Setup

popm is written in python3 and uses the mesa ABM framework with the mesa-geo mapping extension for its interactive mode. Use of a virtualenv or similar is recommended, but not essential.

For batch mode, a more efficient (non-mesa) implementation is used that has no explicit timestepping.

NOTE: the police force dataset used by the model is sensitive data and is not stored in this repo. You will need to obtain a copy of this file in order to run the model.

Native

Firstly, install dependencies

pip install -r requirements.txt

or for conda environments:

conda env create -f conda-env.yml

Interactive Mode

To run a graphical web-based front end:

mesa runserver .

which automatically opens a browser window

Batch Mode

Running batches

For a non-browser batch run for multiple scenarios and/or more detailed data collection:

python -m batch_simple <scenario-file>

Scenario files are kept in scenario

Model output files are saved to a named subdirectory of model-output

As some scenarios contain contain 10000+ runs, batches can be parallelised using MPI (on a suitable platform). On ARC4 use the shell scripts provided, e.g.

qsub mpi_batch.sh 19-small

or for those with only a few possible combinations, just run serially:

qsub batch.sh 1-large

Postprocessing batch output

To summarise runs (by force) use e.g.

python analysis/by_time.py 19-small-noall

which produces output like:

requirement_frac    mean-0.1 mean-0.4 mean-0.6 mean-1.0
location
Avon and Somerset      3h41m    6h24m    12h5m   17h45m
Beds Cambs Herts        1h0m     8h0m    16h0m   17h35m
Cheshire               1h44m     4h0m    4h44m    16h0m
City of London         1h11m    1h11m    1h11m    16h0m
Cleveland              2h51m    5h51m    5h51m    16h0m
Cumbria                2h50m    5h50m    7h10m   16h40m
Derbyshire             2h15m    5h15m    5h15m    16h0m
Devon and Cornwall      4h0m    4h58m    8h48m   16h38m
Dorset                 2h11m    5h11m    9h11m   18h18m
Durham                 2h59m    5h59m    6h35m   16h24m
...

Troubleshooting

If the dependencies don't install, giving errors like undefined symbol: Error_GetLastErrorNum it may be because mesa-geo package has an external (non-python) dependency on (e.g.) libspatialindex-dev which can be installed with e.g. apt.

Profiling

(Mesa is not the fastest...)

Install kcachegrind and pyprof2calltree, then profile using cProfile e.g:

python -m cProfile -o profile.log batch.py scenario/test.json scenario/test.csv

And visualise like so:

pyprof2calltree -k -i profile.log

Or just use pstats for a simpler text visualisation (without call graphs)

Data Sources

• Boundary data: http://geoportal.statistics.gov.uk
• Population data (6/2019 is the latest complete dataset): https://www.ons.gov.uk/peoplepopulationandcommunity/crimeandjustice/datasets/policeforceareadatatables