1
$\begingroup$

I'd like to detect changes in mean in time series data. I've been using R successfully with the ecp (cp3o) and BreakoutDetection libraries. I need to transfer my code from R to Matlab and are therefor looking for similar, or any, change detection methods so I don't have to start from scratch.

I've gone through Matlabs toolbox library and havn't found anything so far. The data could be seen as periodically stationary with noise. The changes will probably be larger than 2 standard deviations from the mean. Outliers should be minimal and can mostly be removed in preprocessing.

I'd be thankfull for any tips on how I could do the change detection quickly and easy with Matlab.

Kindly Gustaf

$\endgroup$
  • $\begingroup$ Have you considered CART models? $\endgroup$ – EngrStudent - Reinstate Monica Sep 15 '15 at 17:59
  • $\begingroup$ I havn't but I think it could work. I'll give it a try. I implemented CUSUM which was very easy :) and it works quite well for what I need. $\endgroup$ – GustafG Sep 29 '15 at 14:32
  • $\begingroup$ I like that there are AICc formulations for CART, so there is some information theory indicating when a transition has occurred, even if your data is sparse. $\endgroup$ – EngrStudent - Reinstate Monica Sep 30 '15 at 1:54
  • 1
    $\begingroup$ @EngrStudent do you have any links for CART models in this context? $\endgroup$ – dr.blochwave Oct 1 '15 at 18:43
  • $\begingroup$ @blockwave - stacks. do you have a preferred language or architecture? MatLab, java .. other? $\endgroup$ – EngrStudent - Reinstate Monica Oct 1 '15 at 18:44

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.