0
$\begingroup$

I have a dataset of certain user activity per week (e.g. purchasing an item or using a service per week) for the past 52 weeks and for 100K+ users. The matrix is very sparse (85% of the entries are zeros). so, it looks something like this:

          W52  W51 .... W01
User01     0    1        0
User02     0    0        0    
...
User99     0    0        2
...

W01 is the most recent week.

My question is whether there are good techniques to cluster the users based on their corresponding time series.

$\endgroup$
  • 1
    $\begingroup$ Not sure if this Dynamic Time Warping together with Time Series Distances would help, but there is a relatively new package in python you can try: github.com/wannesm/dtaidistance, and this one a rather older example of Dynamic Time Warping together with KNN: github.com/markdregan/… $\endgroup$ – TwinPenguins Aug 23 '19 at 5:27
  • $\begingroup$ I think your data is too coarse for any meaningful results. How do you intend to verify the patterns you find? $\endgroup$ – Has QUIT--Anony-Mousse Aug 25 '19 at 7:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.