Time Series in Enterprise Miner - invalid time interval

[nolanmb]

Was trying to get a forecast from Enterprise Miner to predict sales/sku for a particular store.  My dataset has 3 columns, Date, SKU and Sales Units.  Once I put the TS Data prep node I keep getting the error "invalid time interval".  I've tried automatic, daily, weekly but no good.  My dates are daily from 2017/1/29 to 2018/1/18.

Any suggestions would be greatly appreciated.

[TomKari]

Are your dates in SAS date format? If not, you probably need to convert them.

Tom

[nolanmb]

Exactly how do I do that?  Is it something I need to do at the csv file level?  When I bring it into EM I had to edit the variable to tell it's role was a Time ID but that didn't seem to do the trick.

I'm a student and only have basic experience (but did get 2nd place in our Kaggle competition! using EM.  An R program beat me).

What I'm really trying to do is work on forecasting our case project - data = unit sales, date, store# and sku#.  We have 1 years worth of actual data.  I thought SAS Forecast Server would be a great tool to use and got our professor to get me access to the on demand version.  Alas, it looks like you can't upload your own data with the "on demand for academics" version so that was a bust.  So I wondered if EM couldn't do something.

Everyone else is doing it in R but I wanted to try and do it in SAS.  We have 27 million rows of data!

If you could be as specific as possible since I'm a newbie!

[TomKari]

Wow! Way to go! I'm very impressed by your accomplishments.

Before I can give you any advice, some questions:

1. Are you bringing in your data from a .csv file? If so, EXACTLY what format do the entries for your dates have in the .csv file?

2. What SAS tools besides EM can you use? The instructions will vary depending on which tool you are using. Tools that would be useful for what you need to do would be SAS DM (base SAS), SAS Studio, or Enterprise Guide.

Tom

[nolanmb]

Ok, figured out my date issue.  Focusing on the 80/20 rule and looking at only trying to forecast 11 of the 1500+ stores.  Forecast studio has brought up a summary page but the "producing forecasts" has been running for over an hour.  Is that normal?

All I'm hoping to get out of this at the end of the day is a "pick list".

The "R" folks are nipping at my heels.