1
$\begingroup$

I am learning machine learning from scikit-learn and reading its docs.

Clustering clusters groups based on the Euclidean distance and filters them by different ways ex: Gaussian distribution, or mean-shift...etc.

But none of the clustering algorithms cluster samples based on the variation ratio.

EX: below every items has there sold numbers of everyday.

Item,D1,D2
A,1,5
B,10,50
C,4,70

The variation ratio below:
A:500%
B:500%
C:1750%

So A and B are the same group, C isn't.

Are there any clustering algorithms that can cluster time series dataset based on variation ratio (or quantity)?

$\endgroup$

1 Answer 1

0
$\begingroup$
  1. Extract features such as variation ratio
  2. Cluster the extracted features instead of the raw data
$\endgroup$
1
  • $\begingroup$ That's a good solution. I found another way is to scale the value of features to [0,1]. Then cluster it. It can reach good clustering results. $\endgroup$ Commented Jan 28, 2019 at 8:19

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.