1
$\begingroup$

I have the following data:

  • Over 50 activities (e.g. sleeping, eating, studying, watching TV)
  • Observations of many people from ages 40-90 on how many minutes they spend on each activity per day
  • For each age, I have the average amount of minutes spent on that activity, so that I have the pattern of time spent for each activity across time

If I plot my data it looks something like the following, where each line represents one activity and the x-axis represents age intervals: enter image description here

Is it possible to cluster these lines in terms of their pattern to see if any activities follow the same trend over time?

$\endgroup$
2
  • $\begingroup$ Thoughtfully select a distance measure between your lines (profiles, series). Apart from simple distance like euclidean there exist special ones (such as McCrae index etc) specifically to compare "profiles". There are also special distances designed for time processes. After you sevt what suits you, do a cluster analysis on the matrix of distances. $\endgroup$
    – ttnphns
    Commented Mar 21, 2019 at 7:34
  • $\begingroup$ I agree with Anony that there are hardly at all any clusters of patterns seen on your pic, ar least at a first glance $\endgroup$
    – ttnphns
    Commented Mar 21, 2019 at 7:37

1 Answer 1

2
$\begingroup$

Yes.

Obviously you can dump the data as k-dimensional vectors for example in k-means.

Just the quality will likely be quite low. Preparing it carefully can help, but given the plot I'd assume there just is not a lot of pattern in there.

$\endgroup$
3
  • 1
    $\begingroup$ Thanks, this is not what the actual data looks like - this is just a sample chart of what the chart would be like (not in terms of patterns) $\endgroup$
    – Grint
    Commented Mar 22, 2019 at 14:05
  • $\begingroup$ Asking us to suggest ways to cluster random other data won't help you with your real problem much, usually. Data can be very different, and there isn't the one universal approach. $\endgroup$ Commented Mar 22, 2019 at 18:36
  • $\begingroup$ thanks, this is actually a subset of the data I have. I didn't include all the lines since it would be super messy $\endgroup$
    – Grint
    Commented Mar 24, 2019 at 17:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.