[OSGeo Africa] Points Clustering/ Grouping - Seeking a Solution

[E. Ndava]

Hi Gavin.

Great to hear from you too.

Looks like it's a matter of trying several implementations, as you put it.
No silver bullet.

I will dig deeper on DBSCAN and give another go at k-means and brave it out
with GRASS.


*"I don’t think any method will give you the three pre-determined fixed
number clusters it seems you are looking for."*
Something close to start with is acceptable and can be modified with some
manual intervention. My dataset is small and the desired clusters few. Now
I see trouble brewing with a larger dataset, more clusters and weighted
points.

Thanks again

On Wed, 31 Aug 2022 at 07:08, Gavin Fleming <gavinjfleming at gmail.com> wrote:

> Hi Eric
>
> Nice to hear from you!
>
> also try DBSCAN
>
> k-means and DBSCAN are available natively in QGIS, PostGIS and GRASS, so
> also try these in the different implementations
>
> GRASS also has OPTICS and a couple of others (in v.cluster)
>
> I don’t think any method will give you the three pre-determined fixed
> number clusters it seems you are looking for.
>
> hth
>
> Gavin
>
> On 31 Aug 2022, at 00:15, E. Ndava <erick.ndava at gmail.com> wrote:
>
> Hello List Members,
>
> I have a set of 64 points I intend to group by proximity into three
> clusters (of 22, 22, 20) and am looking for a QGIS, PostGIS or any other
> OSS solution to this problem.
>
> I have Googled for one on the web and have skimmed
> https://gis.stackexchange.com/ and have not found an appropriate one
> (a.k.a ran out of patience).
>
> My starting point - Constrained K-Means Clustering, in QGIS did not
> yield acceptable results [
> https://spatialthoughts.com/2021/01/31/equal-sized-kmeans-qgis/ ]
>
> Looking forward to some pointers.
>
> Erick.
>
> PS:
> List Inappropriate Content:  (Ignore)
>
> The below clusters, Red, Green and Blue were produced by esri software
> (source under image)
>
> <group_analysis_attempt_2.PNG>
>
> - I also tried the esri way (ArcGIS), Sort Tool which too did not produce
> acceptance (i.e. clusters not equally sized),  results but something close.
> Source - [
> https://community.esri.com/t5/spatial-data-science-questions/group-every-near-50-points-together/m-p/167184?commentID=687372#comment-687293
>  ]
>
> - I do not have access to ArcPro.
> _______________________________________________
> Africa mailing list
> Africa at lists.osgeo.org
> You can UNSUBSCRIBE at https://lists.osgeo.org/mailman/listinfo/africa
>
>
> _______________________________________________
> Africa mailing list
> Africa at lists.osgeo.org
> You can UNSUBSCRIBE at https://lists.osgeo.org/mailman/listinfo/africa
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/africa/attachments/20220831/a63a1203/attachment.htm>