0
$\begingroup$

What are the commonly use techniques to cluster function ? In my case it is time series data from different subject. My time is discrete (1, 2 till 24) and my response is continuous.

$\endgroup$
  • $\begingroup$ Do you have continuous or discrete time? What is the range? Reals or integer? $\endgroup$ – Karel Macek Nov 3 '17 at 6:06
  • $\begingroup$ @KarelMacek i have updated my question to answer your comments thank you $\endgroup$ – Wis Nov 3 '17 at 6:08
  • $\begingroup$ Is it fair to say that loosely "functional data" is time-series data? $\endgroup$ – Vladislavs Dovgalecs Nov 3 '17 at 16:23
0
$\begingroup$

For clustering, we need to measure the distance of the functions. There are some standard metrics in functional analysis based on some norms, e.g. P norms like $$ \int_X ||f(x)-g(x)||dx $$ check this https://en.wikipedia.org/wiki/Lp_space#The_p-norm_in_finite_dimensions . The trick is that P norms for discrete cases (and your data belong to that) correpond to standard Manhatan/Eucliean etc metrics that are commonly used in clustering of all arts such as k-means or hierarchical.

$\endgroup$

Your Answer

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

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