[OGRUG] Multi-agent modeling including geospatial data

[Joseph Potvin]

Thanks Paul,

Certainly there's been a lot of progress in Netlogo + R (Netlogo is
GPL licensed)
http://cran.at.r-project.org/web/packages/RNetLogo/index.html
http://finzi.psych.upenn.edu/R/library/RNetLogo/tutorial/Thiele_jss.pdf

I'd hoped there was also R integration with Breve or RepastPy but I've
not been able to locate any particular integration with those two as I
have see so far with RNetlogo.  I'd prefer to do my work with
R&Python, not R&Netlogo since I have later application extensibility
I'd like to leave as straightforward work in Python.

Joseph


On Thu, May 29, 2014 at 11:28 AM, Paul Gilbert <pgilbert902 at gmail.com> wrote:
> R has several useful tools for dealing with and plotting geospatial data.
> There may be speed considerations around data extraction, local calculation
> or server, network speed, etc, as been discussed in the "Interactive
> geospatial data with Leaflet ..." thread. Specific code examples can make it
> a bit easier to focus on speed issues.
>
> On the other part of Joseph's question, regarding multi-agent modelling, I
> am not aware of good solutions in R. (That does not mean they don't exist
> and I would be interested to hear if anyone knows of something.) Multi-agent
> modelling covers a lot of things and this is not my field, but in the
> examples I have seen there are two sorts of problems that make R less than
> ideal. The first is that the problems do not easily "vectorize", which means
> speed is a problem. Typically, even with compiled languages, speed is an
> issue for these models and people are looking toward multi-core, cluster, or
> GPU processing. The second is the need for a "language" to define the model.
> These languages tend to be somewhat specialized to the problem domain. Once
> you decide on a language to define the model, the underlying computer
> language for doing the computation is largely hidden.
>
> My approach to this would be to first consider the many already defined
> languages for multi-agent modelling (see e.g.
> http://en.wikipedia.org/wiki/Comparison_of_agent-based_modeling_software),
> and then look at whether an interface from R to one of these would be useful
> for doing plotting, summary statistics, etc, with the existing modelling
> software handling the main computation. I have done this sort of thing with
> some econometric modelling software and it can be relatively simple. R is
> fairly good at interfacing to other things.
>
> Paul
>
> On 05/27/2014 07:00 AM, Amin Adatia wrote:
>>
>> I am interested in this as well. The kind of plot I looking to build
>> is for distribution of Health Effects (Disease, Mortality, etc) at
>> Locations and with a Time spread so we can see the progress or lack
>> thereof in the application of Health Services; For example, Cancer
>> Incidences over time and at various locations. This to perhaps
>> identify the effect of the environment and/or Treatments.
>>
>>
>> Regards Amin Adatia KnowTech Solutions Inc ( http://knowtech.ca )
>> Mobile +1-613-864-8378
>>
>>> --Forwarded Message Attachment-- From: jpotvin at opman.ca To:
>>> r-ug-ottawa at r-project.org Date: Mon, 26 May 2014 10:36:49 -0400
>>> Subject: [OGRUG] Multi-agent modeling including geospatial data
>>>
>>> I am currently planning a multi-agent model that will include
>>> geospatial input and output data. I have not done this before and
>>> would be interested in some orientation from someone who is
>>> experienced with the appropriate R packages.
>>>
>>> -- Joseph Potvin
>>> jpotvin at opman.ca
>>>
>



-- 
Joseph Potvin
Operations Manager | Gestionnaire des opérations
The Opman Company | La compagnie Opman
jpotvin at opman.ca
Mobile: 819-593-5983