Release 3.40.04 Highlights
- New feature: Allow PMML scoring code exporting.
- New feature: Provide processing time estimate in categorical variable auto-grouping, in the case of large number of categories.
- New feature: Any grouped value can be ungrouped back to the original values.
- New feature: Auto-grouping can be done with user-specified number of groups and 'similarity' (by predictive pattern) among variable predictive values.
- Enhancement: ATS functions are not built for excluded variables, for increased processing speed, but they can be built if the exclude checkbox is unchecked
07/08/2010
Gains# 3.50.03 released. Key: unlimited categorical conversion... (detail on right)
03/15/2010
Gains# 3.40.04 released. Key: PMML exporting... (detail on right)
Release 3.50.03 Highlights
- Enhancement: any numerical-type variable can be converted to categorical variable by its distinct values. By default, if number of distinct values from a numerical-type variable is 100 or less, Gains# automatically converts it to categorical type.
- Enhancement: c-statistics for model evaluation is now using much better algorithm to signficantly reduce the computational time.
- Issue fix: in very rare situations, clicking certain areas in Explore tab before ATS transformation may trigger an 'Unhandled Exception' error message. This version fixes the issue.
Terminology note:
With a consideration of the underlying technology and our desire to make it simpler for the new users, the new version adopts the following changes:
- ADR is changed to ATS (Adaptive Transformation Score) - this reflects the key technology behind Gains#, which is a flexible, adaptive transformation from raw data value to a consistent, comparable score scale.
- GPS to Predictive Strength - This is the overall variable predictive value. We think the Pred. Strength is a self-explanatory term and easier for the new users.
Copyright Ó 2001-2009 InfoDecipher Corp. All Rights Reserved.