The Publish Analysis delivers the SQL and parameter information needed to deploy analytic models. This is done by making available for execution the SQL that scores the model and the SQL that builds the Analytic Data Set (ADS) that is input to the model. It can deliver Scoring SQL, ADS SQL or both. The Publish Analysis writes information to metadata tables in Teradata that are read by the web-based Model Manager application (a component of the Teradata Warehouse Miner product). It can also be used to deliver information to the AppCenter web-based application that provides functionality similar to that of the Model Manger.
When published to the AppCenter, the SQL representing a model score and/or analytic data set becomes an “App”, with substitution parameters available for such things as output table names and Variable Creation target dates and literal parameters, as applicable. When connected to a Teradata database, the user may publish either to the Model Manager application or to the AppCenter application. When connected to an Aster database, the user may only publish to the AppCenter.
One of the inputs to a Publish analysis is the analysis to be published. It is selected via a drop-down box on the analysis parameters tab of the INPUT panel. This analysis may be anything in the ADS, Reorganization or Scoring categories with the exception of a Refresh or Publish analysis, or it may be a Free Form SQL analysis. If the analysis to be published takes input from another analysis, the SQL generated by the published analysis and the entire chain of referenced analyses is published.
While analyses in the Descriptive Statistics or Statistical Tests categories cannot be published directly, the SQL can be captured and placed in a Free Form SQL analysis and published from there, provided the SQL creates a table or view when publishing to the Model Manager (not validated). Note, however, that the SQL generated by the Algorithms in the Analytics category may not be published.