Logistic Regression - Class Imbalance

Question

Is there any technique included in SSAS 2008 for handling class imbalance? Like the Fraud example from  "Data Mining with SQL Server 2008", my data set has a very small percent of cases with the class label 1 (around 5%), the rest have label 0.

Is there any technique native to SSAS that will help weight cases? Or do I have to resort to oversampling?

Thanks,

Matt

Answer 1

There are no native techniques.

---

Tatyana Yakushev [PredixionSoftware.com]

Answer 2

Thanks Tatyana.

I am seeing all of the predicted probabilities fall between a range of .50 to .99. Would this be consistent with a dataset with such a severe class imbalance?

Thanks,

Matt