Visual Forecasting Events

[Ghabek]

Hello

Im in VIYA visual forecasting 8.4 and using a dataset to forecast shipments during covid time - data is form Jan-Jun2020 weekly and I want to inject covid events called

covidonset and covidpeak using the code form users guide of VF......Im attaching the sas program here and sample of the dataset used to forecast....I get an error in pipeline saying the name of the event field needs to be CHAR but it is.....dont know what is going on.....

any help to get it working in VF would be appreciated

[solarflare]

Hi Ghabek,

I noticed your covid.sas file is missing a semicolon on the last line.

Try fixing that and rerunning it. I was able to create the event table on VF 8.5 after that.

[Ghabek]

i fixed that but i get the error I sent you in the screen shot for the
event table column not being char......anyway we can hop on call quick?

[solarflare]

hmm, let me investigate a little more first. I'm less familiar with changes in events between 8.4 and 8.5.

[Ghabek]

this is the error o get in pipeline for autoforecasting

[image: image.png]

[TammyJackson]

I can't see the image.

[Ghabek]

here is the log file.....

[TammyJackson]

I can't see the log file, either.

You might find this article helpful:

https://communities.sas.com/t5/SAS-Communities-Library/Improve-Accuracy-by-Adding-Events-in-SAS-Visu...

[Ghabek]

i mainly want to prepare the forecast for a second covid wave in Aug and Sep....do i need to create a new event var called covidwave2 from Aug-sep30??

my data ends Jun 8......

thanks

[TammyJackson]

So an event produces a dummy variable that models an effect such as an outlier, level shift or ramp. Modeling the effect improves the model. Often the event is modeled separately, and then the seasonality and the trend are easier to model once the event effect is removed. If an event occurs in history and in forecast, then the impact of the event on history can be used to forecast the future effect of the event. I would suggest first trying to determine the historical effect. Is that effect a temporary level shift? a ramp? You can try different effects, different dates, or even several different effects and then combine the effects. Once you know the historical impact, you might be able to guess better what the future impact will be.

[TammyJackson]

In order to use the information obtained from the historical effect on the forecast, you would just use a single event with some timing values (dates or datetimes) in the historical period and some in the forecast period.

[Ghabek]

here is the log file......

cant attach it...can i have your email and i can send it?

im a sas partner