Which clustering algorithm would you use, moreover which distance measure, in case of analysis in frequency domain?

I would like to perform Discrete Fourier Transform on time series and perform clustering, but I'm not sure what to use in case of frequency spectra. I could use K-means with Euclidean distance as distance measure, but on the other hand I could use K-medoids and a different distance measure. I'm not able to find any information what is usually done in practice in such a case.

| cite | improve this question | | | | |

If you consider/convert the frequency spectrum to be a kind of histogram, all the distributional and divergence measures can be used. Also, quadratic forms and earth mover's distance (EMD) may be an option. They can't be used with k-means, but with many other algorithms.

| cite | improve this answer | | | | |

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.