I am building a model for time series data that contains ~3 distinct patterns: stable at a 35 deg C, small oscillations around 35C, and large oscillations not centered around 35 deg C. Here is a photo:

enter image description here

A first modelling attempt involved creating classes based off daily st.dev, and constructing random forest based CART algorithm from there. However, given the time series structure, I dont think this is the best idea for real-time detection.

What would be good algorithms to look into?


  • $\begingroup$ stats.stackexchange.com/questions/332475/… might help. The issue is simple the soultion can be complex. There are 4 distinct issues at hand whivh deal with the expected value and the covariance structure of residuals. Time varying parameters and time varying error variance are important considerations. Primitive solutions (regardless of how cheap they are ) often suggested here to identify models by those who should know better do not work when things get complicated AS they nearly always do..Post your data and I will try to embellish.. $\endgroup$ – IrishStat Mar 13 '18 at 20:53
  • $\begingroup$ @IrishStat can we move this to a chat? chat.stackexchange.com $\endgroup$ – Evan Mar 21 '18 at 21:46

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.