1
$\begingroup$

Trying to apply seasonal_decompose on timeseries data. It looks something like this:

            modal_price
Period  
2014-11-01  1469
2015-01-01  1258
2015-03-01  1112
2015-04-01  1373
2015-06-01  1370
2015-07-01  1406
2015-08-01  1520
2015-09-01  1860
2015-10-01  1436
2015-11-01  1455

Using freq=11 and seasonalcomponents are all Nan. What should i do in order to detect the seasonality type(multiplicative/additive) of this data?

$\endgroup$
  • $\begingroup$ Is that sampled data? Typically, you cannot have missing data for time series analysis. But to your question, there are methods for determining seasonality, but perhaps it might help to understand what language and packages you want to use. $\endgroup$ – Skiddles Nov 1 '18 at 14:43
  • $\begingroup$ Hey! There were missing vakues but I sampled it. I now have a consistent dataset. I use python and can use any package, just looking for a solution here. $\endgroup$ – Vipul Rustagi Nov 2 '18 at 15:04
2
$\begingroup$

Once divide the time-series by its Moving Average and once subtract the moving average from it.

If the seasonality is additive, then the result of subtract will have almost similar fluctuations in magnitude.

If it is multiplicative, then the division result has such a property.

Like the figures below from my course in Feature Engineering:

enter image description here enter image description here enter image description here

So the seasonality was multiplicative as the division has similar magnitudes of fluctuations.

Please note that more linear the dynamic of the time-series is, better this naive approach works.

Hope it helps!

| improve this answer | |
$\endgroup$

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.