"Predictable Column Name" and "Predictable Value"

Hi,

I tried to utilize Mining Accuracy to analyze my models.

Mining Model Predictable Column Name Predictable Value
--
NaiveBayesModel
DecisionTreeModel

When I want to choose an option for "Predictable Column Name" on NaiveBayesModel row or DecisionTreeModel row, there is no option/value/choice on the drop box. There is also no option/value/choice for the "Predictable Value" column.

When I clicked "Lift Chart" tab to see the accuracy chart, it gave me this error message: "No mining models are selected for comparison."

The models are as follows:

(

[CustKey] KEY,

[Gender] TEXT DISCRETE,

[BikeModels] TABLE PREDICT_ONLY

(

[Model] TEXT KEY

)

)

Please assist!

Mary

The accuracy charts provided in the product only work for scalar predictable columns. In your model, the target is the BikeModels nested table, which cannot be handled by the tools|||This is because you don't have any predictable case level columns and the accuracy charts don't support nested table outputs. Since the prediction in your case is a nested table containing multiple bike models, there's no standard way to compare the accuracy of two models.|||

Ok,

If there is no standard way for predicted columns in a nested table, is there any non-standard way or perhaps using DMX query without relying the tool to measure the accuracy of the models? Is there any online resources that discuss more about this issue?

Mary

|||

I posted this response to a similar question - hope this helps. http://forums.microsoft.com/MSDN/ShowPost.aspx?PostID=110238&SiteID=1

The issue in accuracy regarding nested tables is in the definition of the question. "Accuracy" means that the model predicts correctly what you are asking. With nested tables, it's not entirely clear what an "accurate" prediction would be. For instance, it could be that a recommendation occurred within the top n answers, or it could be that the predicted value for a particular nested key was correct, etc.

Let us know if the linked response helps, or if there are other areas of "accuracy" you want to explore.

Thanks

-Jamie

|||

Yeah, it helps.Thanks!

Mary