Creating clusters based on attribute and proximity [duplicate]

Question

In my map as you see below, I have my population by district.

I want to create 50 clusters with similiar sums of the population and by proximity. How I can achieve this?

For each cluster, I want the districts are next to each other (As you can see in yellow in the picture).

Have you taken a look at the various tool which pop up when you enter cluster in the search window of the tool box? — Erik, Commented Apr 5, 2019 at 6:16
yes, it seems they all used clustering with values but without considering the spatial proximity... Like this one : ekazakov.info/projects/abc_tutorial.html — Simon GIS, Commented Apr 5, 2019 at 6:20
Yes, but will now not considerate the population factor. SO DBSCAN will only clusters in 50 base on distance... — Simon GIS, Commented Apr 5, 2019 at 6:48
@SimonGIS were you able to solve your problem ? Would you mind sharing how ? — YeO, Commented Jun 29, 2021 at 8:46

MartinT · Accepted Answer · 2019-04-05 05:36:03Z

Not a full-fledged answer, but:

Build an adjacency matrix of your polygons, to know which is adjacent to which. This may be done either through the R provide scripts, or Python, as far as I know there is no current implementation built in QGIS;
Search via breadth-first search for the neighbours, and keep on checking whether you got to the population required.

You will need tweaks - you do not say if the population is given by a constant, or determined as a fraction of the overall population.

1 Answer 1