IBM® SPSS Modeler scoring adapter installation

For some databases it is possible to enable SQL pushback of the majority of the SPSS® Modeler model nuggets. In this way, model scoring can be performed within the database, avoiding the need to extract the data before scoring. This pushback can either use the native SQL within SPSS Modeler or, where available, use additional SQL scoring adapters that are tailored for different databases.

The scoring adapters support scoring of most model nuggets in a stream, with the following exceptions:

• Association Rules (with list data), TimeSeries, Sequence, PCA, STP, and TCM are not supported.
• Association models do not support transactional format.
• Text Analytics (TA) - Field mode is supported in Db2 LUW but not the other databases. When using the TA scoring adapter in field mode, the returned row length may contain many scoring output columns, which can result in some database limits being exceeded. For example, you might see the error message TOO MANY ITEMS RETURNED IN SELECT OR INSERT LIST. Sometimes these limits can be changed in the database but, if that is not possible, the recommended approach is to use record mode to score the model.

The use of scoring adapters that enable data to be scored by the generated models within the database to avoid data transfer. SPSS Modeler enables integration with IBM and non-IBM databases, and enable models to be deployed faster and with greater efficiency.

Where scoring adapters are installed into the relevant databases the SQL generation option generates scoring adapter SQL by default, unless you specifically choose to override this. The databases for which scoring adapters are available are:

• Netezza
• Teradata
• Db2 LUW