[Buildingdata] Re: outliers

[Colin Williams]

Chris, 

I am using some code to do something similar on opendatamap, specifically for the food hygiene maps as they contain outliers. It's a bit of a hack, using some random sampling, but it's quick and may be useful to you. I can send more details later. 

Christopher Gutteridge <cjg at ecs.soton.ac.uk> wrote:

>I was thinking of doing something like that for this map: 
>http://maps.southampton.ac.uk/service/?src=http://id.southampton.ac.uk/dataset/places/latest
>
>which is rather screwed up on centering since we opened a building in 
>Malaysia...
>
>On 16/04/2013 10:59, Andy Turner wrote:
>> This may be equivalent to your horrible hack, but one can truncate
>(as 
>> in a truncated mean statistics) as follows: Calculate the centroid of
>
>> all the points, calculate the distance to all points from this 
>> centroid, discount say the 5% of the furthest away points,
>recalculate 
>> the bounding box (or circle or whatever you are using) for the 
>> remaining points and use this. This could be computationally slow if 
>> you are doing it lots of times, and setting the threshold is 
>> arbitrary, but this could work. There are various ways to speed this 
>> up, for instance you could simply use the centroid of each of the 
>> buildings rather than each of their complete geometries. HTH Andy
>>
>>
>> On Tue, Apr 16, 2013 at 10:29 AM, Christopher Gutteridge 
>> <cjg at ecs.soton.ac.uk <mailto:cjg at ecs.soton.ac.uk>> wrote:
>>
>>     So I've got a situation. On our main campus, the database lists
>two
>>     buildings as part of that site, although they are quite a way off
>the
>>     main site and really screw up any automatic map renderings as
>they
>>     cause
>>     the important part to shrink when scaled-to-fit.
>>
>>     The ideal solution is to have better data, but I am not allowed
>to
>>     edit
>>     what estates send me, although I can augment it.
>>
>>     I currently remove them with an awful hack in the code, but what
>I'm
>>     thinking is creating a relationship between a thing and seomthing
>it's
>>     listed as having within, but should not be included on the
>default
>>     map. eg.
>>
>>     <http://id.southampton.ac.uk/site/1> hasOutlier
>>     <http://id.southampton.ac.uk/building/1580> .
>>
>>     Any thoughts?
>>
>>     --
>>     Christopher Gutteridge -- http://users.ecs.soton.ac.uk/cjg
>>
>>     University of Southampton Open Data Service:
>>     http://data.southampton.ac.uk/
>>     You should read the ECS Web Team blog:
>>     http://blogs.ecs.soton.ac.uk/webteam/
>>
>>     _______________________________________________
>>     Buildingdata mailing list
>>     Buildingdata at ecs.soton.ac.uk
><mailto:Buildingdata at ecs.soton.ac.uk>
>>     http://mailman.ecs.soton.ac.uk/mailman/listinfo/buildingdata
>>
>>
>>
>>
>> _______________________________________________
>> Buildingdata mailing list
>> Buildingdata at ecs.soton.ac.uk
>> http://mailman.ecs.soton.ac.uk/mailman/listinfo/buildingdata
>
>-- 
>Christopher Gutteridge -- http://users.ecs.soton.ac.uk/cjg
>
>University of Southampton Open Data Service:
>http://data.southampton.ac.uk/
>You should read the ECS Web Team blog:
>http://blogs.ecs.soton.ac.uk/webteam/
>
>
>
>------------------------------------------------------------------------
>
>_______________________________________________
>Buildingdata mailing list
>Buildingdata at ecs.soton.ac.uk
>http://mailman.ecs.soton.ac.uk/mailman/listinfo/buildingdata

Colin Williams
crw104 at ecs.soton.ac.uk
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ecs.soton.ac.uk/pipermail/buildingdata/attachments/20130416/711c7909/attachment.html