3
$\begingroup$

I have approximately 174 univariate time series that I would like to forecast. These are all country observations that have been thoroughly cleaned with no outliers or missing values.

I would like to fit each time series to the appropriate forecasting method with "enough care" that I can. This may be a little laborious.

Can I speed my forecasting selection/ by checking if similarities exist with different countries? And from that stage choose appropriate forecasting techniques for each clustered group of countries?

The end goal is not accurate forecasting but rather trend analysis

I have read from here that Dynamic Time Warping may be the optimal time series clustering solution.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

Similarities between countries would be based upon examination of models thus build the models first.

Trend analysis is at best a vague concept . Trends can be deterministic or stochastic (as part of an arima model). Either trend detection needs to be concerned with level shifts ( which are not trends ) . see ML preprocess to achieve stationarity and stochastic vs deterministic trend/seasonality in time series forecasting

If you wish you can post your data and I will try and provide more details.

Improve this answer
$\endgroup$
3
  • $\begingroup$ hi @irishStat, agree on the trend analysis being vague! Can you clarify on your advice to "build models first"? If I single out countries that have different trends than I would be able to treat X amount of countries with the same forecasting model. From that point, I would be able to fit the model loosely to the similar features of each cluster. $\endgroup$
    – editworthy
    Jan 8, 2020 at 15:37
  • $\begingroup$ "singling out countries that have different trends " requires model building or very rough (inadeqiate) eye judgement . Objectively better to construct models that separate signal from noise and then go on to classification ... either formally or judgementally.; The modelling approach is to form a SARMAX model for each series. search on "user:3382" sarmax to see some of my previous reflections $\endgroup$
    – IrishStat
    Jan 8, 2020 at 17:09
  • $\begingroup$ I found your post here. Thank you for your help I will research this from here. For any viewers of this question, I will come back with my concluded solution. $\endgroup$
    – editworthy
    Jan 9, 2020 at 9:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.