Introduction to agent-based modeling

GIS extension

So I am going to talk a bit about the 
NetLogo GIS extension today, right.

The model that I want to talk about is in
the models library under IBM textbook

chapter eight. And is it called Netlogo
ticket or it is called ticket sales

right.

And this was inspired by the dissertation 
of Tseng, who was a student at 
Maryland while I was there.

And she was interested in, in working 
with her dissertation chair Wendy Moe 
in trying to understand how live event
ticket sales affected each other.

This model never quite got to the point of 
really kind of exploring that question in 
detail from an agent-based model (???)

Again her dissertation is much better at 
that, but it is not agent-based.

But it does examine, it is a first 
attempt at trying to get to that point, 
right

So, in this model, you will notice that 
it takes a little while to set up and run

And that is because what we are going 
do, is we are actually going to do is

pull in data about real geographical
spaces.

And this was data that was obtained 
from the Federal government

the US Federal government's census. 
Basically drawn from Tiger which is

their tool that provides free data about
the census to anyone who goes and 
logs in essentially .

Now some of the data you have to pay for 
and stuff like that but

much of it is free.

So what I have done is I pulled in a map
of Manhattan with the census boundaries

laid out on it. and then I populated
each of those census boundaries

with a representative agent for every
household, actually I think it is like one

for every ten households that are in that
particular census tract.

So, it is kind of interesting just what
you see from this right away.

For instance, there is no one living in 
Centreal Park, which runs right down the

middle as you might expect, right. And 
then, what I also have is a venue

that I have kind of located in the world,
And what this allows you to do is then

explore how people decide whether or 
not they are going to this venue over that

venue. and using real demographic data
about that particular area.

So, in this case, right all of these agents 
decide that is is worth their time and

money to attend.

And you can kind of understand

how placing different venues

might have. and eventually what you'd like to have

two venues that compete against each other

and maybe provide discounts and offers

and things like that. That's how the model runs.

Let's pause for a second and I'm going to show

you how the code actually works.

So, in order to read in a NetLogo extension,

you have to start the model code with extensions and

the name of the extension, and you saw this

a little bit in unit 4 where we went over how

to build the viral marketing model.

And so in this particular case, I'm going to read in the

GIS extension. and that GIS extension is one of the ones

that is built into the NetLogo package when you download it.

and the first thing it does is call this setup maps command, and the

setup maps command actually says 'load using the GIS command dataset, load these shapefiles

from a ???

and it's going to put those into variables.

And shapefiles, for those who've never played

around with them before, are a GIS

way of describing polygons, points and lines

and how they describe the space of the world.

So this describes the roads and the tracks so you can see

what's going on. The next thing we have to do, is

we have to actually, since the way GIS works is that you all this data about the whole world

.... but you only want to look at a particular point part of it. cos you

don't want to look at the whole world when we study Manhattan

so, what we're going to do is, we're going to set the world

what's called the envelope (the space wrapping around the world) and

wer='re going to set that envelope to the union of the roads dataset and the census tracts dataset.

These have specified envelopes, so there's something that sets

the outer limits of the roads and the outer

limits of the census tracts. So we're going to take that and we're going

to use that to create the union of our system and create the envelope.

And then we're going to display the roads and display the tracts

and we're going to essentially do those by drawing

setting the drawing colour for GIS to red,

and tell NetLogo to draw it with a line thickness of 1.

And what that does, is actually draw the red lines

that you see back here. And in fact we can

turn that off and on because of the way

we specified that code.

Now this is in the drawing layer

and so there are few ways to interact with it

we can display the tracts and illustrate them as well

and we'll show you how to do the interaction in just a bit

so one of the things we do, and this is where we

do the interaction, is we say 'for each of the tracts

ask the patches that are intersecting with that tract

to set their tract ID in the patch to the GIS

property STFID, which is this tract ID value

that is stored in the tract dataset.'

so now that we've done that, each patch knows which tract

it is attached to in the GIS dataset

and when we load the patch data, we're going to

the households file, which contains the population

for the households in each of the tracts

but the most important thing is that we read in that household population

and it turns out this GISID2 specifies the census tract that that

population is describing. So we can ask the patches

that have the tractID that matches that census tract to

set their population to read from the string of that population.

Now there's one small problem with that, there could be

multiple patches in each census tract and we don't want to have

multiple accounting by having the population in each of these.

So the next line deals with that. It says ask the patches with

population > 0, set their population to the population divided

by the count of the patches which have the same tract ID.

So if I have 2 patches that are in that tract because

they have the same tract ID then I actually split the population

between the two of them.

And so, now, using this final intersection, we've intersected

the drawing layer in terms of the tracts with the actual patches.

It's this GIS-intersecting that is determining that for us.

Then after that, the rest of the code is pretty much standard

agent-based modelling code, it's just going to

determine the consumers are from the venue,

and then make a decision about whether to buy

based on that.

So, that's it for the GIS extension

in the next talk, I'll talk about the network extension