0

I am looking for an algorithm to cluster timestamped events using the weighted elapsed time between the events as the only dimension. There are different types of events. There are constraints saying that some events cannot occur before other events. This is an indication that the events should be in different clusters (regardless of their distance). Does anybody have any idea of what approach to use?

Magnus Eklund
  • 101
  • 1
    Seems like a time series clustering problem. You can try using _dynamic time warping (DTW)_ for that - check [my relevant answer](http://stats.stackexchange.com/a/131284/31372). Not sure about constraints, though. – Aleksandr Blekh Jan 16 '15 at 14:23
  • 1
    Thanks, but I ended up with a totally different approach. I constructed a graph connecting all non-conflicting events. I then used (some) of the maximum cliques of the graph as my clusters – Magnus Eklund Jan 28 '15 at 22:32
  • You're welcome. I'm glad that you've found the solution. – Aleksandr Blekh Jan 28 '15 at 22:37

0 Answers0