I want to cluster time-series concerning sales of products. In the database I have 26weeks after launching each products and units sold each week.

One of the method of clustering is to cluster parameters of growth curve fitted to time-series.

Can somebody tell me what should I do before fitting logistic curve to time-series? Should I use standardization, create model to each time series, use lsm to count parameters and then cluster them?

I would be grateful for any explanation.


Can I ask what you want to use the information for? If you are trying to disseminate information, parameter estimates from logistic regressions can be difficult to convey.

But let me see if I am following what you are asking:

  1. you have n products of which you are looking at the likelihood of being purchased.

  2. you have 26 time points where you want to fit a time series that fits the data best (rather than say, a prediction model like an ARIMA).

  3. After you have fit a unique model (such that each model has its own set of predictor variables which might or might not be unique to that model) to each product, you wish to find some unity to the disparate models by clustering.

Let me know if I am on the right track with what I am assuming and then I can let you know what I do as prep for longitudinal logistic regressions.


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.