Step 2 clustering demand

3 posts Page 1 of 1
Mark
Posts: 5
Joined: Tue Nov 21, 2017 1:48 pm


After getting passed step 1 executed twice for using a layer of address points and for using OSM to get the roads, I have encountered several problems.

1) With GNI Planner 1.0 I was able to do all 4 steps but when I measure lengths in QGIS, they do not equal the lengths in the GNI tables. Measuring with Google, I get the same lengths as with the QGIS tool. - SOLVED, somehow I changed something and now the lengths are correct

2) Step 2 clustering demand. I installed GNI Planner 3.0 to calculate a small town in the USA (2,563 addresses). I am using EPSG 4326 and 3825. When I try to calculate clustered demand, the algorithm runs for at least an hour at process v.net.connect and in the end, there is no clustering because each demand point is in group 1. Using K-Means it also runs for at least an hour but I do get 20 clusters. The clustering runs much faster in GNI Planner 1.0 (less than a minute) so I tried to use GNI Planner 1.0 for step 1 and GNI Planner 3.0 for step 2 but that did not work, even when changing the demand field to dtype. What caused this and what is the solution?

3) Where do I find the lengths from the final nodes to the address points?

I had these questions under step 1 GPON demand but demand clustering is a new topic. And I need eliminate my other username.
gni_support
Posts: 75
Joined: Wed Sep 20, 2017 1:05 pm


Hello,

2) I do not recommend working with the spatial system of coordinates WGS 84, because calculating the length on the ellipsoid really takes quite a long time.
In the GNI Planner 3.0 version the default coordinate system is ETRS89/Poland CS92. If you want to work with this version of the Planner FTTH, you need to change the coordinate systems of all layers and adjust settings : Processing (Menu bar) -> Options -> Providers -> Gni Planner FTTH -> Intermediate layers crs

3) The route of connecting the address point, i.e. the length from final node to the address point, depends on the designer.
Mark
Posts: 5
Joined: Tue Nov 21, 2017 1:48 pm


Many, many thanks for this bit of information, which I tried out immediately but am getting to around now to write a response. I implemented a different CRS for my project and it worked right away! The clustering process can do 2,541 data points in 10 seconds. Excellent.

"3) The route of connecting the address point, i.e. the length from final node to the address point, depends on the designer."
But could this calculation from FINAL node to building point be easily implemented or would it be possible to get the lengths going past the houses in the streets? That means, from the FINAL node, what is the length to pass the houses but not necessary connect them?
3 posts Page 1 of 1
ekskluzywne stoły stoły włoskie nowoczesne