Which parameters for the bundling

[lhos...@gmail.com]

Dear group members,
I have succeeded in importing my data in jflowmap. Great.
These are daily migration data from mobility survey in Nord-Pas-de-Calais
region, in northern France.
I want to bundle the edges in order to read the data and see the underlying
spatial organisation, and to identify places and flows for structuring a
car-sharing service. Crystal clear to everyone? Right.
My problem is to find the right parameters for the bundling of edges.
so far I have obtained very nice but completely irrelevant images.
trying one by one playing with the parameters just worsens my problem.
that is why I need help!
I provide all the files in the message. they just contain the number of
persons travelling between each communes in the region in a regular weekday.
Thanks in advance,
Alain

[Ilya Boyandin]

Hello Alain,

Using a smaller "step size" and a higher "step damping factor" can work with this dataset (see the picture below). But honestly, I don't think that bundling is really useful here. The movements seem to have a star-like structure with a few hubs attracting people from the surrounding areas This results in a quite a clean structure already, in my opinion, which bundling can actually misrepresent.

Best regards

Ilya

--
You received this message because you are subscribed to the Google Groups "jflowmap" group.
To unsubscribe from this group and stop receiving emails from it, send an email to jflowmap+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Alain L'Hostis]

Dear Ilya,
thanks so much for your help!
indeed my ultimate goal is to bundle very tiny flows in order to help designing a car-pooling/car-sharing system so I want to see were can some very tiny flows meet and bundle, so so far I still hope your tool can help me.
My problem currently is the following:
I used your parameters:
default except
Step size: 0,001 (instead of 1)
Step damping: 0,75 (instead of 0,5)
Steps in 1st cycle: 50 (instead of 100)

And the image I got is not the same as the one you got. I atach it to the message.
I've tried with Ubuntu and Windows with the same unsatisfactory result
I used jFlowMap version 0.16.6

Thanks in advance,
Alain

[Ilya Boyandin]

Yes, it's is different because I used only top 1000 flows to make it faster.

Try something like this with the full dataset:

--

[Alain L'Hostis]

Dear Ilya,
just to let you know that I have tried another software with an edge bundling algorithm (tulip) and the results are  better from my point of view. Or do you think one could find some parameters in jflowmap that could match this type of image?
In case you would like to test tulip it I provide you with the tlp file.
thanks for all your efforts,

Alain

Le vendredi 9 mai 2014 16:33:56 UTC+2, Alain L'Hostis a écrit :

[Ilya Boyandin]

Hi Alain,

Thanks for letting me know! Good that it worked for you.

Regards

Ilya

--