turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Data Mining
- /
- SAS Enterprise Miner - fit statistics for scored d...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

05-27-2015 06:07 PM

Hello,

I created a gradient boosted model in SAS Enterprise Miner, and I've also imported a new data set into the program to score using the resulting model. I know how to generate a list of predicted values for the target variable for each entry in the data set. However, I'd also like to compute fit statistics for the new scored data as I can get from the model comparison node for the training and validation data set. Does anyone know what I could enter into the SAS code node to generate the same set of fit statistics from the scored data set?

Would be really appreciative of any help.,

Accepted Solutions

Solution

05-29-2015
02:43 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to tonysmith1

05-29-2015 02:43 PM

It sounds like the Model Import node is the way to go. And you can use it one of two ways - I think the first is how you would use it, since it sounds like you scored the data set already. And I'm assuming you have the target in the data you want to score or already scored.

1. Create a data source for the scored data that you created, then connect IDS->Model Import, with Import Type=Data. Run, and you will get the usual model fit statistics. You can then connect that to a Model Comparison node to get more assessment statistics.

2. If you haven't already scored, you can register the model from Gradient Boosting (either use the Register Model node or right-click on the node and select Create Model Package, then register it from the Model Packages folder in the Project Panel), then create a data source for the data you want to score, set up a flow IDS > Model Import, with Import Type=Registered Model and the model you registered selected for Model Name.

Hope that helps.

All Replies

Solution

05-29-2015
02:43 PM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to tonysmith1

05-29-2015 02:43 PM

It sounds like the Model Import node is the way to go. And you can use it one of two ways - I think the first is how you would use it, since it sounds like you scored the data set already. And I'm assuming you have the target in the data you want to score or already scored.

1. Create a data source for the scored data that you created, then connect IDS->Model Import, with Import Type=Data. Run, and you will get the usual model fit statistics. You can then connect that to a Model Comparison node to get more assessment statistics.

2. If you haven't already scored, you can register the model from Gradient Boosting (either use the Register Model node or right-click on the node and select Create Model Package, then register it from the Model Packages folder in the Project Panel), then create a data source for the data you want to score, set up a flow IDS > Model Import, with Import Type=Registered Model and the model you registered selected for Model Name.

Hope that helps.