2

I have a question about performing clustering with clouds of points in which one dimension - representing time - is somewhat protected.

To make it super clear, consider this video

cloud_of_points

With a naked eye one may see some dense clouds flying around like mosquitoes, they may represent several things entering and leaving a scene. Now suppose we have an array of 3-dimensional points (x,y,time) and apply some naive clustering (say DBSCAN)

clustering

Now the clustering is quite good, except that meeting events are considered in the same cluster, coming up with X-trajectories. Now if there was some way to treat the third coordinate differently, perhaps one may recover the ground truth. Which algorithms may be well suited for this problem?

Has QUIT--Anony-Mousse
  • 76,138
  • 12
  • 138
  • 194
user8440809
  • 21
  • 2

1 Answers1

1

There exist density-based stream variations of DBSCAN.

They are supposed to solve exactly your scenario of clusters moving slowly as time advances, along with clusters appearing and disappearing.

I'm not convinced by all these "streaming" methods though. They don't seem to use any real data, only simulated and non-natural streams like "poker hands".

Has QUIT--Anony-Mousse
  • 76,138
  • 12
  • 138
  • 194
  • Is there code for these algorithms available online? Perhaps implemented in libraries like SKLearn? I would like to test it on my data. Thank you very much! – user8440809 Aug 10 '17 at 07:55
  • Have you looked into the R package `stream`? The examples in the documentation look very similar to the examples above. – Michael Hahsler Aug 10 '17 at 09:33