In the first KJDM URI example, a “select” statement can be
directly used as a dataset to build or apply models, or compute
statistics. This feature is particularly interesting when, for example,
there is a need to filter lines or join information from different tables.
The second example shows direct access to a table or a view prepared
in advance. The user and password can be passed directly in the URI,
but could also be assigned as default values in the odbc.ini parameter
file to cope with some security issues. The access to all major data-
bases from Windows and UNIX platforms is supported.
A specific extension can be associated with KJDM to read also
proprietary formats of most commercial data mining vendors.
The only extension that KXEN has implemented concerns the
available languages in which classification, regression, and cluster-
ing models can be exported. The model resulting from an ExportTask
execution will work on the data provided in the same format as for
the BuildTask (which means unprepared data).
The possible exported formats are listed here:
• SQL for MySQL
• SQL optimized for SQL Server
• SQL optimized for Teradata
• SQL for IBM DB2 and Oracle
• UDF for Oracle
• UDF for IBM DB2
• UDF for SQL Server
• UDF for Teradata