1
$\begingroup$

I'm trying to generate a one-year forecast for the number of units sold by a retail company. I'm using monthly data from 2017 and 2018. The forecast is for 2019, and I'm using the data from the months that have already passed (January-July) as test data. So far, I have had good results generating the forecast for every category using an innovations state space model for exponential smoothing as described by Hyndman et al. (2018)

The problem is that I just came across a category (let's call it category X) where the number of units sold had a significant overall drop since January due to a change of suppliers. I believe this drop will continue during the rest of the year due to business reasons. Data looks like this.

Time series for category X

Is there a way I can incorporate this information into the innovations state space model? Should I use a different method to forecast category X? Any suggestion is welcomed.

Reference:

Hyndman, R.J., Athanasopoulos, G. (2018) Forecasting: principles and practice, 2nd edition, OTexts: Melbourne, Australia. OTexts.com/fpp2. Accessed on August 2019.

$\endgroup$
  • $\begingroup$ What model are you using? ETS? If so, you cannot include variables in such a model. An ARIMA can do it. $\endgroup$ – user2974951 Aug 29 '19 at 6:13
  • $\begingroup$ Yes, I'm using ETS. I'll check out ARIMA models then. Thank you. $\endgroup$ – MarianaMG2205 Aug 29 '19 at 17:03

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.