Angoss recently released an end to end scorecard building tool –ScorecardBUILDER – based on the same underlying workflow capabilities released in Angoss KnowledgeSEEKER 9.0 (reviewed here). This product is designed to automate or otherwise streamline time consuming manual steps in the process of building a predictive scorecard.
As noted, ScorecardBUILDER is based on the new workflow-centric look and feel Angoss has recently released. This is extended in a number of ways for the ScorecardBUILDER tool:
- There’s a helper for adding optimal binning and weight of evidence for a dataset. Each variable can then be assessed, visualized and optimized so that it is monotonic – trending in a single direction – and predictive. This helper creates the variables using a Weight of Evidence transformation.
- A new logistic regression node is added for use in the workflow. This builds a stepwise regression model from selected variables. The user can force certain variables to be used and easily include only the variables that came from the weight of evidence analysis. Correlation between variables and other potential issues can be identified. Once the model step is done the standard workflow nodes are used to validate and analyze the model.
- Finally the new scorecard node takes the logistic model and scales it. The node allows the user to specify the base score, the rate of increase/decrease based on changing odds, the target category that gets the higher score and more. The result is a scorecard with variables, bins and cumulative scores.
The resulting scorecard can be viewed and edited in the tool, reported on and exported as SAS/SQL code, or copied to MS Office. The scorecard could also be used to score external datasets within the ScorecardBUILDER software.
Angoss is one of the vendors in our Decision Management Systems Platform Technology Report and you can get more information on ScorecardBUILDER here.