Release Notes - IBM SPSS Modeler 15.0 Fix Pack 2

Content

Installing
Windows 8
New features
Scripting compatibility with previous versions
Variable file parsing
Entity Analytics Configuration Guide
Known Issues

Installing IBM SPSS Modeler 15.0 Fix Pack 2

For information on downloading and installing the fix pack, see the relevant Download Document. Depending on your installation you can chose from the IBM SPSS Modeler Professional components and the IBM SPSS Modeler Premium add-ons.

Windows 8

Support for Windows 8 has now been introduced for Modeler client products. However, the 15.0.0.0 versions of Modeler Professional Administration Console, Modeler Premium Administration Console, Modeler SNA, Modeler Entity Analytics Unleashed, and IBM SPSS Data Access Pack (SDAP) products will not install on Windows 8.

To install these products, download and install the 15.0 Fix Pack 1 versions of them from Fix Central and install directly before applying the FP2 version. See the Fix Pack 1 Release Notes for further details.

If you require the SDAP to install on Windows 8, please contact IBM SPSS Technical Support.

New Features

New Database Support
In this fix pack, support has been added for the following database: DB2 LUW10.1.

Cognos 10.2 Support
In this fix pack, support has been added for Cognos 10.2.

SAP HANA and IBM DB2 on UNIX or Linux
Connecting to SAP HANA, or IBM DB2 using the IBM DB2 ODBC driver, from Modeler Server in a UNIX or Linux environment requires Modeler Server to be configured to use a new ODBC wrapper, "libspssodbc_datadirect_utf16.so". To configure Modeler Server to use this wrapper, complete the following steps:

1. Navigate to your Modeler Server "installation directory/bin" folder.
2. Remove the existing link file "libspssodbc.so" using the following command:
   rm -fr libspssodbc.so
3. Link the new wrapper to "libspssodbc.so" using the following command:
   ln -s libspssodbc_datadirect_utf16.so libspssodbc.so
4. Configure your Modeler Server startup script "modelersrv.sh" to source the IBM SPSS Data Access Pack odbc.sh environment file by adding the following line to "modelersrv.sh":
   . /pathtoSDAPinstall/odbc.sh
   where <pathtoSDAPinstall> is the full path to your IBM SPSS Data Access Pack installation.
5. Restart Modeler Server.
6. You must add the following parameter definition to the DSN in your odbc.ini file to avoid buffer overflows during connection:
   DriverUnicodeType=1

The wrapper "libspssodbc_datadirect_utf16.so" is different from the wrapper detailed in the Configuring ODBC drivers on UNIX systems section of the Modeler 15 Server for UNIX Installation Instructions. If you try to use these ODBC drivers with Modeler Server configured as documented in the Modeler 15 Server for UNIX Installation Instructions you will see the following behavior:

• Table names will not be displayed correctly in the database source node.
• All string data extracted from tables will be displayed incorrectly.

Note: The "libspssodbc_datadirect_utf16.so" wrapper is also compatible with the other Modeler Server supported ODBC drivers.

Teradata 14
The Teradata ODBC driver is the only Teradata driver supported with this version of Teradata server. Consequently, access to Teradata 14 on SuSE11 zLinux is not supported because the Teradata ODBC driver does not support this platform.

Node Updates
The following nodes have been updated:

• Binning Node

An option has been added to the “Fixed-width binning” mode that causes the same bin intervals to be used for all binned fields. When connecting to a remote server, this option is ignored unless the Fix Pack has also been applied to the Modeler Server installation.
The behavior of the binning method "Equal Count", when handling ties using "Keep in Current", has been refined so that the number of records are distributed more evenly among percentile groups, especially for the first group.

• Auto-Classifier and Auto-Numeric Model Builder and Model Applier Nodes

An option has been added that generates the individual model outputs, in addition to the output of the model ensemble.

• Evaluation Node

An option has been added that overlays profit criteria for all chart types, where previously this option was only available for the Profit chart type.

Scripting Compatibility with Previous Releases

There is an error in the Scripting and Automation Guide, which states that model nuggets may now be inserted in the stream automatically if automatic model replacement is enabled in the stream and the stream is executed by a script. The scripting behavior in this area has not changed and models need to be inserted into the stream using the "insert" scripting command, as in previous releases. For example:

for n in 1 to 4

execute :neuralnetworknode

insert model realincome:applyneuralnetworknode connected between :type and :sample

execute table

delete realincome:applyneuralnetworknode

endfor

Variable File Parsing

When using the Variable File node, the check box "Line delimiter is newline character" is clear by default. With this setting any blank values at the end of data rows are retained in the data during processing, as they are in Modeler 14.2. If you require blank end of row values to be removed when processing your data, select the "Line delimiter is newline character" check box. Modeler 15.0 stripped blank end of row values even when this option was not selected, making it's behavior inconsistent with Modeler 14.2.

Entity Analytics Configuration Guide

The fix pack includes an update to the EA Configuration Guide. If you are using Entity Analytics with Collaboration and Deployment Services, Decision Management, or Solution Publisher you should follow the instructions given in this Configuration Guide.

Entity Analytics on Solaris

For Entity Analytics to work on Solaris, the minimum system patch level is Solaris 10 Update 6.

Entity Analytics - using an existing repository

If you want to continue using a repository created prior to applying the fix pack, you need to complete the following steps to upgrade the schema of the repository to the fix pack level.

From a Command Prompt:

cd <modeler-install-directory>\ext\bin\pasw.entityanalytics

updateschema.bat|.sh <username> <password> <repository-name> <Optional - repository-location>

Note: The Optional repository-location is only needed if you have changed the default location of the repository.

Modifying pwutil on UNIX

If Modeler Server 15 on UNIX is configured to use the DataDirect Driver manager, then pwuitl needs to be aware of this in order to run correctly. If you have not configured Modeler Server to use the DataDirect driver manager with the SDAP you will not see this issue.

If you are experiencing this issue then after installing the IBM SPSS Data Access Pack and configuring Modeler Server to use the DataDirect driver manager, edit the pwutil script in the Modeler Server bin directory and add the line:

. <full path to the odbc.sh within your IBM SPSS Data Access Pack installation>

so if you had installed the IBM SDAP into /usr/SDAP the above line would look like this:

. /usr/SDAP/odbc.sh

Please ensure you add this additional line immediately before the last line in the pwutil script that executes the pwutil_exe binary (that is, after the existing definitions and exports of the library path variables that are in the pwutil script by default).

Known Issues

Storing or Configuring Text Analytics Nodes or Streams into Collaboration and Deployment Services Environment
If you are using C&DS 4.x, an error occurs when storing or configuring Text Analytics nodes or streams. To resolve this issue, carry out the following instructions.

Add the path <C&DS_installation_root>/components/modeler/ext/bin/spss.TMWBServer to the system path environment variable used by the application server running the Collaboration and Deployment Services.

The system path environment variable differs between platforms:

• On Windows the environment variable is PATH
• On AIX platform the environment variable is LIBPATH
• On other Unix platforms the environment variable is LD_LIBRARY_PATH
  

The location of the environment variable differs between application servers:

• For WebSphere, the environment variable can be set through the admin console for the appropriate server.
• For JBoss, the environment variable can be found in the wrapper.conf file.
• For Weblogic, the environment variable can be found in the domain’s start script.

On HP-UX platform, make sure that all the files in the directory <C&DS_installation_root>/components/modeler/ext/bin/spss.TMWBServer/bin have execute permissions.

Note: <C&DS_installation_root> refers to the full installation path for Collaboration and Deployment Services server. For example: /usr/IBM/SPSS/Collaboration and Deployment Services/4.2/Server

This issue does not affect users of C&DS 5.0.

Champion Challenger
If you are using Entity Analytics, the champion challenger comparison process does not work.

C&DS Context Root
Modeler 15.0 Fix Pack 2 does not support the C&DS context root feature.

C&DS Nicknames
C&DS nicknames are not supported by the Entity Analytics repositories in Modeler 15.0 Fix Pack 2. When naming EA repositories, a naming scheme should be used so that the repositories are given unique names.

DB2 CLOB type
Modeler 15.0 Fix Pack 2 does not support the CLOB type in DB2.

Properties for the Cognos Connection using Modeler Batch
When using the property: cognos_connection in Modeler Batch (clemb), the string value of the property should be escaped correctly, as required in the Java application command-line arguments.

For example:

clemb -stream batch.str -P:cognosimportnode.cognos_connection={"Cognos_server_URL", login_mode, "namespace", "username", "password"} -execute

This string should be escaped as:

clemb -stream batch.str -P:cognosimportnode.cognos_connection={\"Cognos_server_URL\",login_mode,\"namespace\",\"username\",\"password\"} -execute

In addition, if the user name in the property is, for example: ibm\user, the string should be escaped as: \"ibm\\user\".

Text Analytics Installation Instructions - Non-Administrator User

In the section titled "Enabling Access for a Non-Administrator User", the following entry includes an incorrect parameter name:

"Modify the line: #tmfc_client_data_dir = c:/data/IBM_SPSS_Text_Analytics_Client to read:
Tmfc_client_data = c:/TextData"

This should read:

"Modify the line: #ta_client_data_dir = c:/data/IBM_SPSS_Text_Analytics_Client to read:
Ta_client_data = c:/TextData"

Scoring a Text Analytics model in Analytical Server shows different results from native Modeler scoring

When scoring Category models in Analytical Server all categories in all records may be incorrectly scored as "F". To prevent this, use Text Link Analysis (TLA) to perform an extraction when you have a TLA model that has categories that have been defined using TLA rules.

Text Analytics Server exception when scoring a TA node with no matching categories in Analytic Server

Scoring fails with the error message: "Server exception: java.lang.IndexOutOfBoundsException Index:0 size: 0"

If you see this message, it means that no category has been matched. To avoid the error, on the Settings tab of the scoring nugget, use "Categories as fields" mode instead of "Categories as records" mode.