Hello,
When producing a (homogeneous) network with temporal dynamics:
For example:
Choose Original Timescale: Standard Periods
Number of time slices: ‘3’
time slices distribution: regular
Overlapping periods: false
sequencing: snapshot
If I have chosen a network of 400 nodes, the top400 nodes are re-evaluated for each of the three time slices, right?
Another possibility would be for the top400 to be evaluated only once over the whole period, for example 1991-2021, and not for each of the three slices ‘1991_2001’, ‘2002_2011’, ‘2012_2021’ . (Then only the proximity values would be recalculated for each period.) But that’s not what the algorithm does, is it?
Thanks in advance for your reply!
Aurélien
Thanks Aurélien for this question.
By default, the Top N nodes are selected independently for each period / time slice, which results in only partially overlapping nodes being mapped.
However, you need to adjust the parameters to match the temporal behavior of the dynamic clustering with what you want to analyze:
- Map exactly the same set of nodes at each time step: activating this option computes the Top N nodes selection once over the entire time period, without considering individual time slices (even though the maps are still generated for each time slice).
- Freeze nodes positions when computing temporal maps: if you need to examine each map for every time slice and compare them, you may want to freeze the node positions.
I hope it helps!
L