Optimal site location based on existing point objects

[bobrandall]

MI-L,

I"m somewhat new to GIS and I've been asked to select the optimal
location for a distribution hub based on 4,000 delivery locations.
I've used the Distance Calculator in the past to find the closest
location to candidate sites. In this case however, we have no target
sites and I have to provide our real estate folks with the place to
start (if I can generate a point I can provide them with the closest
intersection).

Would the average Lon/Lat be what I"m looking for? I've looked
through the MI documentation but don't see another tool that can
create a point object based on existing points.

Once we've identified potential sites I'll use the Distance Calculator
to select the ideal location, just not sure how generate the optimal
location from scratch based on existing delivery locations.

Any suggestions would be greatly appreciated!

Bob

[Uffe Kousgaard]

If you can fly from your hub to the delivery locations and all has equal
weight, yes then it is a simple average you are looking for. If they don't
have equal weight, then it is centre of mass you are looking for.

MapInfo is not really needed for all this - Excel will do.

Regards
Uffe Kousgaard

> --
> You received this message because you are subscribed to the
> Google Groups "MapInfo-L" group.To post a message to this group, send
> email to mapi...@googlegroups.com
> To unsubscribe from this group, go to:
> http://groups.google.com/group/mapinfo-l/subscribe?hl=en
> For more options, information and links to MapInfo resources (searching
> archives, feature requests, to visit our Wiki, visit the Welcome page at
> http://groups.google.com/group/mapinfo-l?hl=en
>

[Bill Thoen]

What's your definition of "optimal"? Are you trying to find the point
that's closest to the most number of points, or one that minimizes the
cost of travel to all points, or what? If your definition of optimal is
the point that is most equidistant to all points, there's a good
exchange on StackOverflow about how to find th ecenter of a set of
lat/long points. You ca also download a 25Mb PDF about the problem. See
http://stackoverflow.com/questions/353684/determine-the-centroid-of-multiple-points

--

*Bill Thoen*
GISnet - www.gisnet.com <http://www.gisnet.com/>
1401 Walnut St., Suite C
Boulder, CO 80302
303-786-9961 tel
303-443-4856 fax
bth...@gisnet.com

[bobrandall]

Hi Uffe,

I created Lon and Lat columns, updated them with the value of
CentroidX and CentroidY and confirmed ("Select Count(*), Lon, Lat from
deliveries group by Lon, Lat order by 1 Desc into Selection") that we
do have coincident points, would this be the "weight" value? The
delivery addresses are all within a large urban metropolitan area.

How would I use MapInfo to calculate the center of mass to create my
point?

Thank you for your help.

Bob

[bobrandall]

Hi Bill,

Great questions, "optimal" at this point would be point closet to the
most number of points as I believe that would take into account the
coincident points.

Thank you for the link to the method for calculating a point that is
most equidistant to all points (although it's a bit over my head, I'm
not sure how I'd do this in MapInfo).

Bob

On Aug 9, 1:32 pm, Bill Thoen <bth...@gisnet.com> wrote:
> bobrandall wrote:
> > MI-L,
>
> > I"m somewhat new to GIS and I've been asked to select the optimal
> > location for a distribution hub based on 4,000 delivery locations.
> > I've used the Distance Calculator in the past to find the closest
> > location to candidate sites.  In this case however, we have no target
> > sites and I have to provide our real estate folks with the place to
> > start (if I can generate a point I can provide them with the closest
> > intersection).
>
> > Would the average Lon/Lat be what I"m looking for?  I've looked
> > through the MI documentation but don't see another tool that can
> > create a point object based on existing points.
>
> > Once we've identified potential sites I'll use the Distance Calculator
> > to select the ideal location, just not sure how generate the optimal
> > location from scratch based on existing delivery locations.
>
> What's your definition of "optimal"? Are you trying to find the point
> that's closest to the most number of points, or one that minimizes the
> cost of travel to all points, or what?  If your definition of optimal is
> the point that is most equidistant to all points, there's a good
> exchange  on StackOverflow about how to find th ecenter of a set of

> lat/long points. You ca also download a 25Mb PDF about the problem. Seehttp://stackoverflow.com/questions/353684/determine-the-centroid-of-m...
>
> --
>
> *Bill Thoen*
> GISnet -www.gisnet.com<http://www.gisnet.com/>

[Uffe Kousgaard]

Just do:

select avg(lon),avg(lat) from deliveries

no reason to run the first select count(*)..... query.

[Timothy Mashford]

The free CrimeStat program discusses several methods for determining a central point:

 

Mean Centre - this is the average X and average Y coordinates, however it is not the mean 'location' overall.

 

Median Centre - the median X and Y coordinates, so that an equal number of points sit on either side of the median centre. This is not skewed by outlying points.

 

Centre of Minimum Distance - the location at which the distance to all other locations is a minimum. This is most likely what you are after? There is no standard formula for determining this; rather, CrimeStat generates a fine grid to approximate the location.

 

 

The locations determined using these methods will likely be quite similar, however there are differences so it depends on how precise you want to be.

 

Tim

[mickgibs]

Wouldn't you wish to consider travel / drive time. You seem to be
assuming that distance is a valid proxy for determining an optimal
location.

Introducing travel time - which I believe Mapinfo can do, will allow
you to consider travel routes, road, highways, etc.

Regards
Mick

[Bill Thoen]

bobrandall wrote:
> Hi Bill,
>
> Great questions, "optimal" at this point would be point closet to the
> most number of points as I believe that would take into account the
> coincident points.

> for x

> Thank you for the link to the method for calculating a point that is
> most equidistant to all points (although it's a bit over my head, I'm
> not sure how I'd do this in MapInfo).
>

So to define that a little more precisely, it sounds like what you're
looking for is the coordinate where the sum of the distances to the
other points is at its minimum. How's your calculus-fu? Off the top of
my head (and I'd have to do some Googling in the math pages on linear
equations, differential calculus, and optimization) I think the approach
would be to set up a series of equations of distance from each point,
add them, take the first derivative of that, set it equal to zero and
solve for x and y. And I think I'd use a tool like Excel if the Solver
can handle it or maybe one of the many math tools out there on the net.

But from your questions and the requirements you've been given, it
doesn't sound like you want to (or need to at this point) get into this
very deeply. The level of precision required doesn't seem too great
either, and you need more math background to do this with any
sophistication, but you probably wouldn't be wrong just eyeballing the
location, and then to keep anybody from asking questions you don't want
to answer, print off a couple of pages of this site
http://www.codecogs.com/reference/maths/calculus/differential/linear_simultaneous_equations.php,
mark up the pages a bit, write a title on top like "Optimal Location
Determination Notes" and leave them out where anybody can see them. If
someone still asks how you determined the location, pick up the pages,
show them to the questioner and say something vague and off-putting
like, "It's a bit difficult to explain in layman's terms..." ;-/)/

--
*Bill Thoen*
GISnet - www.gisnet.com

303-786-9961

[Gentreau]

Ah, the voice of experience, how nice to hear it from time to time ... :)

Bill, is your real name actually Wally ?

-----Original Message-----
From: mapi...@googlegroups.com [mailto:mapi...@googlegroups.com] On
Behalf Of Bill Thoen
Sent: Tuesday, August 10, 2010 5:06 PM
To: mapi...@googlegroups.com
Subject: Re: [MI-L] Re: Optimal site location based on existing point
objects

--

[bobrandall]

MI-L,

Tim, I'll download CrimeStat and take a look, sounds like a great
tool.

Mick, once we have sites identified I plan on performing a network
analysis introducing travel routes/costs, thanks for confirming my
approach.

>> "Optimal Location Determination Notes" and
>> leave them out where anybody can see them.

Bill, brilliant suggestion, most the time I mention GIS around here
eyes glass over.

Amazing forum, thanks once again everyone for the great information
and the many approaches, I"ve learned a lot!

Bob

> to answer, print off a couple of pages of this sitehttp://www.codecogs.com/reference/maths/calculus/differential/linear_...