Data Fit Score

  • Article

The Data Fit Score is the average log probability that the analysis model assigns to a case in the test data

— minus —

The average log probability that a marginal analysis model assigns to a case in the data.

The marginal analysis model is a test model that assumes all properties are independent of one another.

The higher the score, the greater is the accuracy of the model.

See Also

Negative Data Fit Scores for Segment models

Workflow for the Predictor Resource

About the Predictor Resource

Analysis Models

Using Dependency Network View

Using Decision Tree View

Copyright © 2005 Microsoft Corporation.
All rights reserved.