3
$\begingroup$

I am working on a sales forecast right now and I am not sure what type of model to use.

I have 17 Quarters of data(4 Full years + 1 QTR) and I am only looking to forecast 2 quarters into the future at maximum. My data is also very seasonal.

My only experience is using a simple multiple regression model, using the quarter number(to account for seasonality) and the time period as independent variables, which I worry wont be reactive enough to new data.

What model would be ideal for my scenario?

$\endgroup$
  • $\begingroup$ Your approach should consider identifying anomalies (pulses,level shifts,seasonal pulses ) while incorporating memory (arima structure) . Stay clear of any pre-specified form like holt-winters and allow the data to suggest the appropriate model. $\endgroup$ – IrishStat Sep 11 '18 at 19:28
1
$\begingroup$

Your best bet is probably triple exponential smoothing (also called Holt-Winters), which you can use through the ETS() function the the R Forecast Package or with the Statsmodels library in Python.

$\endgroup$
1
$\begingroup$

I would suggest a Holt-Winters approach as well. I just verified that a quick Google search brings numerous references. Feel free to post questions here too if you run into trouble.

Further, I would suggest comparing your forecast performance against a "No-Change" or a "Seasonal No-Change" model to verify you're adding value with your process.

$\endgroup$

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.