PMML to SQL

Generates SQL queries from PMML models. The following model types are currently supported:

• Regression (linear, polynomial, logistic)
• SVM (no output of probabilities)
• MLP
• K-Means Clustering
• Decision Trees (no output of probabilities)
• Naive Bayes
• Rulesets

Ensemble models can be translated but result in complicated joins that might not run efficiently.
IMPORTANT NOTICES:

• The node only supports tables where the columns do not contain spaces. Spaces in the PMML document's fields are converted to an underscore (_).
• Due to several missing functions such as 'POW' SQLite is currently not supported.

The generated SQL runs best on MySQL databases. Other databases may have limitations. Postgres BUG #6359 for example limits the size of the generated SQL so that some large queries will not work. Microsoft SQL Server limits the nesting depth of CASE WHEN statements to 10, so large decision tree models cannot be executed via SQL.

Input ports

1. Type: PMML
   PMML model
   The PMML model that is translated into SQL

Output ports

1. Type: Flow Variable
   SQL flow variable
   The SQL select statement for the scoring
2. Type: Table
   SQL table
   Table containing the generated SQL