Deterministic runs for agent-based model

[hsl-petrhaj]

Don't know if this is even possible. But maybe we should at least determine how many runs are needed to get stable results that can be used officially.

[johpiip]

Our agent model synthetizator relies on random so it will always be stochastic. The random seed is set here:

helmet-model-system/Scripts/parameters/zone.py

Lines 4 to 7 in 9ca9d71

	# Seed number for population attributes:
	# int = fixed seed and same population for each run
	# None = different population for each run
	population_draw = 31

A quick plan would be to:

1. Make a few agent model runs (for example, on test network) to verify that all stochasticity is indeed controlled by population_draw.
2. Once that is verified, make several agent model runs on a real network (for example, 2018) and compare results: agents, tours, stuff on result_summary.txt etc.

However, I am closing this as this is not really a model development issue! You should copy this to Trello on "Ideat" as this is a great idea in model usage context!

[hsl-petrhaj]

done