BookmarkSubscribeRSS Feed
SonyaApple
Calcite | Level 5
Hi There,

I have successfully used Forecast Studio (FS) for many timeseries projects. I have encountered a new issue with the data that I'm forecasting.

I am forecasting for a new product by regions. I am comfortable with the growth forecasts selected for the first 2 years however knowing the evolving industry, in the latter years the trend will not show much growth and at one point should level out (ie. logarithmic type of trend).

The problem is, the models selected do not reflect this as the data actuals will show a growth trend.

1. Is there a way in FS where I can select for models to only choose the group of 'log' types of curve fitting models?
2. Or any suggestions on modeling this data?

The dataset is very large with many products and regions so it would be time consuming to select specifically an appropriate logit model for each timeseries at each level.

Thanks in advance.
2 REPLIES 2
Snurre_SAS
SAS Employee
Hi,

I'm assuming you are using Forecast Server version 3.1.

By default Forecast Studio will identify ARIMA models and exponential smoothing models (ESM). The types of default models can be expanded to include unobserved component models (UCM) as well.

If you want to consider curve fitting models in your project I would recommend that you build your own model repository containing the curve fitting models you would like to consider. The model reposistory can then be referenced inside Forecast Studio.

To build a curve fitting model definition go through these steps:
A. Go into the "Modelling View" and select "Add model" in the list of icons
B. Select "Curve Fitting" as the type of model you would like to define
C. Specify the characteristics of your curve fitting model and click ok.

When you're done building the curve fitting model definitions you would like to consider you need to build the model repository containing the model definitions and a model selection list that points to the individual model definitions. The following assumes some familiarity with SAS code and the SAS Management Console. If you do not have that try contacting your SAS administrator. Follow these steps:
1. Open a separate SAS session where you can submit SAS code from the editor.
2. Create a library reference to a place where you want to store your models and repository permanently
3. In Forecast Studio, in the "Modelling View", first delete all models except your curve fitting models.
4. Then select the "Model Selection list code" icon
5. Copy this code and paste it into the editor you opened in step (1).
6. Change the library references in the code from "work.temp" to the library you defined in step (2).
7. Submit the code.
8. If it runs succesfully close SAS and Forecast Studio.
9. To get access to your new modelreposity in Forecast Studio open the SAS Management Console, import the library and set the appropriate user permissions.
10. At this stage it might be necessary to restart the SAS Analytics Platform.
11. To consider your new curve fitting models in a project create a new project.
12. Under "Forecasting Settings" select Model Generation.
13. Check "Models from an external list" and click Browse.
14. If steps 1-13 have been done succesfully you should now be able to locate your library containing model definitions and the model selection list.
15. You can optionally switch off the ARIMA and ESM specifications.
16. Finalize your project settings and run the project.

Steps 1-16 might look like it requires some effort but it is actually rather quick to setup.

I hope this helps.

Thanks,
Snurre
SonyaApple
Calcite | Level 5
Thanks for the details. This worked and something new I learned.

Thanks again!

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

Register now!

Multiple Linear Regression in SAS

Learn how to run multiple linear regression models with and without interactions, presented by SAS user Alex Chaplin.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 2 replies
  • 885 views
  • 0 likes
  • 2 in conversation