SAS Enterprise Miner 6.1 and SAS Enterprise Miner 5.3 contain the %EMCONVERT macro. You can use %EMCONVERT to migrate data mining projects that were developed and run in SAS Enterprise Miner 4.x (4.0, 4.1, 4.2, or 4.3) for use with SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1. When you convert a SAS Enterprise Miner 4.x project into SAS Enterprise Miner 5.3 project format, the process flow diagrams, logs, output, score code, and assessment results from the SAS Enterprise Miner 4.x project are saved in the SAS Enterprise Miner 5.3 project format. SAS Enterprise Miner 6.1 opens SAS Enterprise Miner 4.x projects that were converted to SAS Enterprise Miner 5.3 project format, as well as projects that were created and saved in SAS Enterprise Miner 5.3.
You use the %EM_CONVERT macro to convert a SAS Enterprise Miner 4.x project into SAS Enterprise Miner 5.3 format. The source code for the %EM_CONVERT macro is distributed with SAS Enterprise Miner 6.1 and SAS Enterprise Miner 5.3. The %EM_CONVERT macro converts a SAS Enterprise Miner 4.x project into a project folder structure that SAS Enterprise Miner 5.3 and later can recognize and open.
You can configure the project conversion macro to create:
SAS batch code is platform-independent, so the project creation batch code will run on any SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 server, regardless of its operating environment.
There is a difference between SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 projects that are created directly by the conversion macro and SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 projects that are created by running the project creation batch code file that the conversion macro saves.
If you use the conversion macro to create a complete project (instead of the batch code file), you can display the transferred SAS Enterprise Miner 4.x results and compare them to your new SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 results. When you run the project creation batch code that the conversion macro saves, the logs, output, score code, and assessment results from the SAS Enterprise Miner 4.x project are not transferred to the converted SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 project.
There are fundamental differences between the project directory structures of SAS Enterprise Miner 4.x and the project directory structures that are used with SAS Enterprise Miner 5.3 and later. SAS Enterprise Miner 4.x uses different strategies to store projects, project metadata, and project data than SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1.
SAS Enterprise Miner 4.x Project Directory Structures
SAS Enterprise Miner 4.x stores project directory structures for both local and client/server projects on the SAS Enterprise Miner client machine. For each project, SAS Enterprise Miner 4.x creates these three subdirectories: emdata, emproj, and reports:
You can use SAS Enterprise Miner 4.x to find your existing SAS Enterprise Miner 4.x project directory and browse its subdirectories. Open the SAS Enterprise Miner 4.x project, select the Diagrams tab of the Project Navigator, right-click on a named project icon, and then select Explore. A Windows Explorer window opens and displays the project folder and its location in your file system. On Windows systems, the default storage location for SAS Enterprise Miner 4.x projects is as follows:
C:\Documents and Settings\YourEMUserID\My Documents\My SAS Files\9.1\EM ProjectsThe EM4 Project Location directory contains the DMP file for the project, as well as the various DMD files that represent each process flow diagram in the project. In addition to the DMP and DMD files, the project directory also contains the emdata, emproj, and reports subdirectories.
The emdata directory contains Emdata librefs and files that are created when you run process flow diagrams in your project. If you are working on a SAS Enterprise Miner 4.x client/server project, the project's emdata files are written to the data directory on your SAS Enterprise Miner 4.x server. If you are working on a SAS Enterprise Miner 4.x project that can run locally (without connecting to the server), then SAS Enterprise Miner uses the emdata directory on your SAS Enterprise Miner 4.x client.
The emproj directory contains Emproj librefs and project files that contain information for each process flow diagram, its nodes, target profiler settings, and various registries. The users subdirectory in the emproj directory contains files that represent the users, if any, who are currently sharing the project.
The reports subdirectory contains HTML report files, if your SAS Enterprise Miner 4.x project contains a process flow diagram that has a Reporter node. Reports that were generated by SAS Enterprise Miner 4.x Reporter nodes are not transferred to converted SAS Enterprise Miner 5.3 projects. However, the Reporter node is reproduced in converted SAS Enterprise Miner 5.3 projects. You can run the Reporter node in your converted project to generate and save SAS Enterprise Miner 5.3 Reporter node results.
SAS Enterprise Miner 4.x projects are self-contained. You can use file utilities such as Windows Explorer to archive or copy SAS Enterprise Miner 4.x projects. The SAS Enterprise Miner project conversion process does not modify source SAS Enterprise Miner 4.x project files. If the project that you want to convert is located on a server that has limited user access and rights (such as a production server), you can copy the SAS Enterprise Miner 4.x project files to a work area and then perform the project conversion on the copied project.
If you want to move the project (via ftp, for example) to a different operating system, first put the converted project into a transport file, such as a ZIP or TAR file that will be recognized by the destination operating system.
SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 Project Directory Structures
Unlike SAS Enterprise Miner 4.x, SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 stores all project information on the SAS Enterprise Miner server. For each project, SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 creates a project directory that has five standard subdirectories: DataSources, Meta, Reports, System, and Workspaces. Each folder is briefly discussed as follows so you can understand the SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 project folder structure and what the folders should contain in a newly converted project:
The DataSources and Meta folders are SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 structures that did not exist in SAS Enterprise Miner 4.x. Converted SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 projects are created with empty DataSources folders. The Meta folder contains assorted project and diagram metadata that is generated by SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 projects. Converted SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 projects are created with empty Meta folders.
The Reports folder is also empty in a newly converted SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 project. The HTML report files that the SAS Enterprise Miner 4.x Reporter node creates are not transferred to converted SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 projects. If your converted SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 project contains process flow diagrams that use the Reporter node, you can run the process flow diagrams in SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1, and then save the Reporter node output as an LST or PDF file. The newly saved LST and PDF files are stored in the Reports folder of your SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 project.
The System folder holds SAS Enterprise Miner project system files, such as project start-up and exit scripts. Start-up and exit scripts from SAS Enterprise Miner 4.x projects are currently not transferred by the SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 project conversion macro. Converted SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 projects are created with empty System folders. If your converted SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 project requires start-up and exit code, you can enter the code when you open your converted SAS Enterprise Miner 4.x project in SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1.
The Workspaces directory contains a separate Enterprise Miner workspace subdirectory for each process flow diagram in the project. Each process flow diagram subdirectory is named with an EMWS prefix. The EMWS directories contain the files that are associated with the process flow diagram. The EMWS directory for a process flow diagram also contains a subdirectory for each node in the process flow diagram. The node subdirectories contain files that document each node's property configuration settings, as well as node output files such as results and SAS logs. In newly converted SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 projects, the Workspaces directory and all of its EMWS subdirectories should contain data.
All SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 projects must be located on the SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 server to function, but the location where you store the project on the server is flexible. If you desire, you can distribute projects across multiple directories on your SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 server. If you want to store your converted SAS Enterprise Miner 4.x projects in the same location as your existing SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 projects, it is easy to find the location of your existing projects on the server. To find the storage location of an existing SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 project, open the project and select the project icon in the SAS Enterprise Miner Project Navigator. The Path property in the project's Properties panel displays the path to the project on your SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 server.
The following requirements must be met for the project conversion macro to complete successfully:
The source code for the project conversion macro is called EM_CONVERT.SOURCE. It is located in the EMCONVRT catalog of the SASHELP library in your SAS installation. The source code for the project conversion macro can found in SASHELP.EMCONVRT.EM_CONVERT.SOURCE.
The %EM_CONVERT macro has the following general form:
%EM_CONVERT(
EM4PROJECT=SAS-Enterprise-Miner-4.x-project-path,
PROJECTPATH=SAS-Enterprise-Miner-5.3-or-SAS-Enterprise-Miner-6.1-project-path,
PROJECTNAME=SAS-Enterprise-Miner-5.3-or-SAS-Enterprise-Miner-6.1-project-name,
CREATE=Y|N (generate-Enterprise-Miner-5.3-or-SAS-Enterprise-Miner-6.1-project-folders-with-results),
FILEREF=fileref-for-Enterprise-Miner-5.3-or-SAS-Enterprise-Miner-6.1-project-creation-batch-code);
where
You might observer minor differences when you compare the project process flow diagrams and results of your original and converted projects. For more information about possible differences, see Project Conversion Differences in Nodes.
The following project conversion code examples convert a SAS Enterprise Miner 4.x project into a SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 project, a SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 project creation batch code file, or both:
The project conversion code examples use the following scenario:
A SAS Enterprise Miner user named Jean, whose SAS user ID is Jean, is migrating a SAS Enterprise Miner 4.3 client/server data mining project that is named JeanEM4Proj for use with SAS Enterprise Miner 5.3 The project JeanEM4Proj contains three process flow diagrams. The files for the project JeanEM4Proj are stored on Jean's SAS Enterprise Miner 4.3 Windows client machine. This is how the JeanEM4Proj project appears in the Project Navigator of the SAS Enterprise Miner 4.3 software:
Jean uses the project properties information in SAS Enterprise Miner 4.3 to determine the location of the JeanEM4Proj project files. Jean knows that project files aren't affected by the conversion process, but Jean prefers to work with a copy of the project in a local work directory. Jean closes SAS Enterprise Miner 4.3, and then uses a file utility to copy the entire project file structure to a work directory in the following location on the client PC:
C:\Public\Converted\Jean\JeanEM4Proj
You must close SAS Enterprise Miner 4.3 before copying the Enterprise Miner 4.3 project file structure to another location. When open, SAS Enterprise Miner 4.3 places locks on some project files that can interfere with file copy operations.
This is how the copied SAS Enterprise Miner 4.3 version of the JeanEM4Proj project folders appear in Jean's work directory:
Opening the JeanEM4Proj directory reveals the JeanEM4Proj.dmp project file and a DMD file for each of the three process flow diagrams that belong to the JeanEM4Proj project.
Jean's Windows PC has network access to a projects folder on the SAS Enterprise Miner 5.3 server. The SAS Enterprise Miner 5.3 server uses the Windows operating environment. The network path to the projects folder that Jean has chosen to use on the SAS Enterprise Miner 5.3 server is as follows:
\\EM5Server\EM5Projects\ConvertedEM4\Jean
If Jean's conversion creates a complete SAS Enterprise Miner 5.3 or later project folder structure, the converted SAS Enterprise Miner project will be named JeanEM5Proj. The target location for the JeanEM5Proj project folder structure on the SAS Enterprise Miner 5.3 server is
\\EM5Server\EM5Projects\ConvertedEM4\Jean\JeanEM5Proj
If Jean's conversion creates a SAS Enterprise Miner 5.3 or later project creation batch code file, the project creation batch code file will be named JeanEM5ProjBatch.sas. The project creation batch code file can be saved to any location, but Jean prefers to save the JeanEM5ProjBatch.sas file to the following project folder on the SAS Enterprise Miner 5.3 server:
\\EM5Server\EM5Projects\ConvertedEM4\Jean\JeanEM5ProjBatch.sas
When the project creation batch code is run from the above location, the project will be created in the following location:
\\EM5Server\EM5Projects\ConvertedEM4\Jean\JeanEM5Proj
Jean organizes the following information before creating the conversion macro code:
Jean is now ready to create the conversion macro code.
Create a SAS Enterprise Miner 5.3 or Later Complete Project Folder Structure
Creating a SAS Enterprise Miner 5.3 or later complete project folder structure is the most direct method for converting SAS Enterprise Miner 4.x projects into SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 projects while preserving past project logs, charts, and assessment results. This is the best method to use when the SAS Enterprise Miner 4.x client that is used to perform the conversion has network access to the designated project storage folder on the SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 server. In this situation, Jean can choose to create the converted SAS Enterprise Miner 5.3 or later project in the designated project storage folder on the server, or Jean can choose to create the converted SAS Enterprise Miner 5.3 or later project in the project conversion work directory on the SAS Enterprise Miner 4.3 client workstation and then copy the complete converted project folder structure to the designated project storage folder on the server. Either choice creates identical results, a converted project that transfers saved project results, charts, and logs.
Jean is creating a complete project folder structure, so the CREATE= macro option is set to Y. No project conversion batch code file will be saved, so the FILEREF= option will not be used. Jean decides to let the conversion macro create the converted project in the designated project storage folder that is located on the SAS Enterprise Miner 5.3 server.
Jean launches a SAS 9.1.3 session on a workstation that has access to both the SAS Enterprise Miner 5.3 server and the SAS Enterprise Miner 4.3 project, and submits the following code to the SAS Program Editor:
/* use x to allocate and include the catalog entry */ /* that contains the conversion macro source code, */ /* then de-allocate x to clean up */ filename x catalog 'SASHELP.EMCONVRT.EM_CONVERT.SOURCE'; %inc x; filename x; /* specify path and name values for placeholder */ /* variables em4project, em5path, em5name */ %let em4project=C:\Public\Converted\Jean\JeanEM4Proj; %let em5path=\\EM5Server\EM5Projects\ConvertedEM4\Jean; %let em5name=JeanEM5Proj; /* submit macro using placeholder variables and */ /* specify CREATE=Y */ %em_convert(em4project=&em4project, projectPath=&em5path, projectName=&em5name, create=Y);After the code runs, Jean uses a file utility to examine the newly converted project in the destination directory on the SAS Enterprise Miner 5.3 server. The converted SAS Enterprise Miner 5.3 or later project folder structure was created in the correct location. Expanding the JeanEM5Proj project folder reveals the standard SAS Enterprise Miner 5.3 or later project directory structure.
Expanding the Workspaces folder reveals the EMWS subdirectories for the three process flow diagrams that belong to the project. Expanding one of the EMWS folders reveals the SAS catalog files for the process flow diagram, data sets that were created in the diagram workspace, and catalog files for the process flow diagram, logs, results, and plots. Each EMWS folder also contains a subfolder for each node in the process flow diagram.
Jean closes the SAS session on the SAS Enterprise Miner 4.3 client. Jean is ready to start a client SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 session to open and run the process flow diagrams in the newly converted JeanEM5Proj project. After running each process flow diagram, Jean uses the SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 Results window to compare the results that were generated by each process flow diagram in SAS Enterprise Miner 4.3 and SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1.
Create a SAS Enterprise Miner 5.3 or Later Project Creation Batch Code File
Creating a SAS Enterprise Miner 5.3 or later project creation batch code file is useful when it is not important to transfer the existing SAS Enterprise Miner 4.x project logs, charts, and assessment results to the converted SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 project.
The SAS Enterprise Miner 5.3 or later project creation batch code file is also useful in situations where network connectivity between the SAS Enterprise Miner 4.3 client and the SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 server is restricted or unavailable.
SAS Enterprise Miner 5.3 or later project creation batch code is small in size, portable, operating environment-independent, and easily transported by file transfer protocol, floppy disk, CD, thumb drive, or similar media. The project creation batch code file contains all of the information that is required to create SAS Enterprise Miner 5.3 complete project folder structures. The tradeoff for the small size and portability of the project creation batch code file is that the SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 projects that they create do not transfer the SAS Enterprise Miner 4.x project logs, charts, or assessment results.
For this example, assume that Jean does not have network connectivity between the existing SAS Enterprise Miner 4.x client and the SAS Enterprise Miner 5.3 Windows server. Jean will generate the project creation batch code file in the project work directory that Jean created on the SAS Enterprise Miner 4.3 client PC.
Jean does not want to create a SAS Enterprise Miner 5.3 project folder structure, so the CREATE= option in the macro will be set to N. Jean wants to save the project creation batch code file, so a FILEREF= argument must be supplied to provide the filename and path specification. After the project creation batch code file is successfully saved, Jean will use media to manually transfer the batch code file to the designated project storage location on the SAS Enterprise Miner 5.3 server, where the code can be run to reconstruct the converted project, without its pre-existing logs, charts, or assessment results.
Jean launches a SAS 9.1.3 session on a workstation that has access to both the SAS Enterprise Miner 5.3 server and the SAS Enterprise Miner 4.3 project and submits the following code to the SAS Program Editor:
After the conversion code runs, Jean uses a file utility to examine the newly converted project creation batch file in the destination directory on the SAS Enterprise Miner 4.3 client. The batch code file, named JeanEM5ProjBatch.sas, was created in the desired location. Jean copies the file to portable media and transports the file to Jean's project folder on the SAS Enterprise Miner 5.3 server:/* use x to allocate and include the catalog entry */ /* that contains the conversion macro source code, */ /* then de-allocate x to clean up */ filename x catalog 'SASHELP.EMCONVRT.EM_CONVERT.SOURCE'; %inc x; filename x; /* specify path and name values for placeholder */ /* variables em4project, em5path, em5name */ %let em4project=C:\Public\Converted\Jean\JeanEM4Proj; %let em5path=C:\Public\Converted\Jean; %let em5name=JeanEM5Proj; /* fileref specifies filename and path for project */ /* creation batch code file */ filename x 'C:\Public\Converted\Jean\JeanEM5ProjBatch.sas'; /* submit macro using placeholder variables, */ /* specify CREATE=N, provide fileref for project */ /* creation batch code file to be generated */ %em_convert(em4project=&em4project, projectPath=&em5path, projectName=&em5name, create=N, fileref=x);C:\EM5Projects\ConvertedEM4\Jean\JeanEM5ProjNote: Project creation batch files can be stored and run in any location that is accessible to SAS. Jean's project folder locations in this example are arbitrary. You can freely specify a custom destination for your own project creation batch files.
Once the file JeanEM5ProjBatch.sas is located in Jean's project folder, Jean uses a SAS 9.1.3 (or later) session on the server to run the SAS batch file JeanEM5ProjBatch.sas. The batch file creates the complete project folder structure beneath the JeanEM5Proj project folder. After the project folder structures are created, Jean can open the newly converted project using SAS Enterprise Miner 5.3.
Create a SAS Enterprise Miner 5.3 or Later Complete Project Folder Structure and a SAS Enterprise Miner 5.3 or Later Project Creation Batch Code File
The %EM_CONVERT macro is typically used either to create a complete SAS Enterprise Miner 5.3 or later project folder structure or to save the SAS Enterprise Miner 5.3 or later project creation batch code file. With the exception of transferring SAS Enterprise Miner 4.x project results (which the project creation batch code cannot do), both macro output types generate the same converted SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 project.
Suppose, however, that Jean plans to deploy the converted SAS Enterprise Miner 5.3 or later project on a remote server by transporting the project creation batch code to the server, and then running the batch code there. Also suppose that Jean would like to preview the converted SAS Enterprise Miner 5.3 or later project folder structures before transporting and running the project creation batch code in its new location on the SAS Enterprise Miner 5.3 server. Jean wants to create both types of output in the work area, preview the created project folder structures to ensure they meet requirements, and then transport and run the project creation batch code in its new location with confidence. The path to Jean's temporary work area is
C:\Public\Converted\EM5Projects\TemporaryTo create both types of output, Jean's macro specifies both CREATE=Y and provides a path and filename argument for the FILEREF= option.
/* use x to allocate and include the catalog entry */ /* that contains the conversion macro source code, */ /* then de-allocate x to clean up */ filename x catalog 'SASHELP.EMCONVRT.EM_CONVERT.SOURCE'; %inc x; filename x; /* specify path and name values for placeholder */ /* variables em4project, em5path, em5name */ %let em4project=C:\Public\Converted\Jean\JeanEM4Proj; %let em5path=C:\Public\Converted\EM5Projects\Temporary; %let em5name=JeanEM5Proj; /* fileref specifies filename and path for project */ /* creation batch code file */ filename x 'C:\Public\Converted\EM5Projects\Temporary\JeanEM5ProjBatch.sas'; /* submit macro using placeholder variables, */ /* specify CREATE=Y, provide fileref for project */ /* creation batch code file to be generated */ %em_convert(em4project=&em4project, projectPath=&em5path, projectName=&em5name, create=Y, fileref=x);Jean submits the macro code to the Program Editor of a SAS 9.1.3 (or later) session on a Windows workstation that can access the SAS Enterprise Miner 5.3 server and the SAS Enterprise Miner 4.3 project to be converted.
After the code runs, Jean uses a file utility to examine the newly converted project folder structure in the project work area.
The Temporary folder and the JeanEM5Proj project structure folders were created correctly. Jean selects the Temporary folder and views the file list to also verify that the project creation batch code file was saved.
Jean copies the project creation batch code file JeanEM5ProjBatch.sas to a portable media device such as a floppy disk, CD, DVD, or thumb drive. The batch code file is ready to be copied to its destination folder on the remote SAS Enterprise Miner 5.3 server, where it can be run to create new project folder structures.
Alternately, Jean could ignore the saved project batch code creation file, and use a ZIP or TAR file utility to compress and save the JeanEM5Proj project folder structure to portable media. The media can be transported to the SAS Enterprise Miner 5.3 or a SAS Enterprise Miner 6.1 server and the files restored (unzipped or untarred) to their original form. SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 can recognize and open the transported project file structure. Unlike a project that is created by running the project batch code file, the transported project folder structure transfers the process flow diagram results from the SAS Enterprise Miner 4.x project.
The following instructions explain how to open a newly converted SAS
Enterprise Miner 5.3 or later project using SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 software. The instructions
use the converted project folder structure
here:
.
Open a SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 session on your SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 server. In the SAS Enterprise Miner introductory window, choose New Project to open a Create New Project window. In the Create New Project window, provide the following information in the following order:
C:\Public\Converted\EM5Projects.
View Em4Results ...
Differences can exist between original SAS Enterprise Miner 4.x and converted SAS Enterprise Miner 5.3 or later process flow diagram results that are not caused by the project conversion process.
The data mining tools that are provided with SAS Enterprise Miner 4.x and with SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 are not identical. The set of data mining tools in SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 is an improved and expanded version of the tools in SAS Enterprise Miner 4.x. Over time, numerous SAS Enterprise Miner 4.x data mining nodes have been refined or redesigned for enhanced configuration and computation. In more than one case, the functionality that was performed by a single node in SAS Enterprise Miner 4.x is performed by two separate, more configurable nodes in SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 . There might also be enhancements and changes in the SAS procedure code that underlies a SAS Enterprise Miner data mining node.
As a result, a SAS Enterprise Miner 4.x data mining project that is converted for use with SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 might contain converted process flow diagrams that do not exactly match the original process flow diagrams. Every effort has been made to preserve the original process flow diagram structures, calculations, and results, but technical enhancements to the software might result in some perceptible differences between the original and the converted process flow diagrams. Converted process flow diagrams might contain additional nodes, substituted nodes, or enhanced nodes. These incremental changes often improve the computational accuracy of the node, so some variation in results might exist between the original and converted project process flow diagrams.
When comparing original and converted project process flow diagrams and their results, you should expect to see differences in the implementation of the following SAS Enterprise Miner 4.x and SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 nodes:
Sample Node
When performing stratified sampling in a process flow diagram, the default settings of SAS Enterprise Miner 4.x and SAS Enterprise Miner 5.x Sample nodes vary. In SAS Enterprise Miner 4.x, you use the General tab of the Sampling node to specify the Sampling Method as Stratified.
Then, you use the Stratification tab to choose your stratification variable from the available nominal and ordinal variables in your data set:
The default setting for the status of SAS Enterprise Miner 4.x class variables in the Stratification tab of the Sampling node is don't use. If no stratification variable status is set to use, then no stratification sampling will be performed. Instead, the Sampling node reverts to simple random sampling, because data cannot be sampled by strata without a declared stratification variable. You must change the status of a class variable from its default setting of don't use to use to perform stratified sampling.
In SAS Enterprise Miner 5.x and SAS Enterprise Miner 6.1, you use the Properties panel for the Sample node to choose Stratify as the Sample Method:
Then, you must open the Variables property window to choose your stratification variable by assigning the Sample Role:
You must assign the Sample Role of Stratification to your desired stratification variable, and you must also assign the Sample Role of None to the remaining variables. The default Sample Role for variables in SAS Enterprise Miner 5.x is Default, which means that if a class variable is not explicitly configured as the stratification variable, the node will use all class (binary, nominal, and ordinal) variables by default to perform stratified sampling. If there are no class targets, then the node will perform random sampling by default.
You must use care to ensure that Sample node settings are consistent in both versions of your SAS Enterprise Miner 4.x and SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 process flow diagrams. Otherwise, the default settings of the Sample node might differ. Different data samples might change the analytical results of your converted SAS Enterprise Miner 5.3 and later process flow diagram.
Data Partition Node
The default settings for stratified data partitioning vary between nodes in SAS Enterprise Miner 4.x and SAS Enterprise Miner 5.3 and later. Variances between the default settings of the Data Partition node in the SAS Enterprise Miner 4.x and SAS Enterprise Miner 5.3 and later process flow diagrams can disturb the data partitioning results in some cases.
When performing stratified data partitioning in a process flow diagram, the default settings of the SAS Enterprise Miner 4.x and SAS Enterprise Miner 5.3 and later Data Partition nodes vary. In SAS Enterprise Miner 4.x, use the Partition tab of the Data Partition node to specify Stratified as the data partitioning method.
Then, use the Stratification tab to choose your stratification variable from the available class variables in your data set:
The default setting for the status of SAS Enterprise Miner 4.x class variables in the Stratification tab of the Data Partition node is don't use. If no stratification variable status is set to use, then no stratification is performed. Instead, the Data Partition node reverts to performing simple random data partitioning, because data cannot be partitioned by strata without a declared stratification variable. You must change the status of a class variable from its default setting of don't use to use to perform stratified data partitioning.
In SAS Enterprise Miner 5.3 and later, you use the Data Partition node Properties panel to choose the partitioning method:
If you specify Stratified as the Partitioning Method, you must open the Variables property window to choose your stratification variable by specifying the Partition Role for the stratification variable:
Assign the Partition Role of Stratification to your desired stratification variable, and then assign the Partition Role of None to the remaining variables. The default Partition Role for Data Partition variables in SAS Enterprise Miner 5.3 is Default, which means that if a class variable is not explicitly configured as the stratification variable, the node will use the class target variable as the stratification variable by default. If no class target is found, all class (binary, nominal, and ordinal) variables will be used to perform stratified data partitioning.
You must use care to ensure that Data Partition node configurations match in both versions of your SAS Enterprise Miner 4.x and SAS Enterprise Miner 5.3 and later process flow diagrams. Otherwise, the default settings of the Data Partition node can differ between SAS Enterprise Miner 4.x to SAS Enterprise Miner 5.3 and later. Differently partitioned data can alter the analytical results of your converted SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 process flow diagram.
Replacement and Impute Nodes
The Replacement node in SAS Enterprise Miner 4.x performs both data replacement and data imputation functions. SAS Enterprise Miner 5.3 and later uses two nodes to perform those functions: the SAS Enterprise Miner 5.3 and later Replacement node performs data replacement, and the SAS Enterprise Miner 5.3 and later Impute node performs data imputation. SAS Enterprise Miner 4.x process flow diagrams that contain a Replacement node, when converted for use with SAS Enterprise Miner 5.3 and later, will contain both a Replacement node and an Impute node in the new diagram.
The order of data replacement and data imputation are configurable in the SAS Enterprise Miner 4.x Replacement node. The order of the Replacement and Impute nodes in converted SAS Enterprise Miner 5.3 and later process flow diagrams depends on how the SAS Enterprise Miner 4.x Replacement node was configured.
The order of data replacement and data imputation in the SAS Enterprise Miner 4.x Replacement node is configured in the General subtab of the Replacement node's Defaults tab. The node performs data imputation before performing data replacement, unless the Replace before imputation check box is selected.
The process flow diagrams here illustrate how the order of replace and impute operations affects the conversion process for SAS Enterprise Miner 4.x process flow diagrams:
A significant difference exists between the SAS Enterprise Miner 4.x Impute node and the SAS Enterprise Miner 5.3 and later Impute node that must be considered when comparing analytical results produced by the two project versions. The Impute node in SAS Enterprise Miner 4.x bases its imputation values on a sample of the submitted data set. The Impute node in SAS Enterprise Miner 5.3 uses the entire submitted data set to base its imputation values on. As a result, the SAS Enterprise Miner 5.3 and later imputation algorithm tends to produce a slightly more accurate variable imputation than an imputation that was produced in SAS Enterprise Miner 4.x.
Ensemble and Group Processing Nodes
In SAS Enterprise Miner 4.x, the Ensemble node performs multiple functions. The Ensemble node in SAS Enterprise Miner 4.x can be used to combine different models. The Ensemble node in SAS Enterprise Miner 4.x can also be used with the Group Processing node to accumulate models such as stratification models and bagging or boosting models. When you convert a SAS Enterprise Miner 4.x process flow diagram that contains an Ensemble node for use with SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1, the new process flow diagram contains either an Ensemble node or an End Group node, depending on whether the original Ensemble node performed model combination or model accumulation.
Using the Ensemble node to combine models can improve the stability of disparate nonlinear models. When the individual models use class targets, the combined model averages the posterior probabilities of the individual models. When the individual models use interval targets, the combined model averages the predicted values of the individual models.
The following example shows an Ensemble node used to combine models in a SAS Enterprise Miner 4.x process flow diagram. The example also shows the process flow diagram after it is converted for use with SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1. The Ensemble node still combines the models in the converted diagram:
.
When you perform group processing in SAS Enterprise Miner 4.x, you can use the Ensemble node to accumulate individual group models. Group processing in SAS Enterprise Miner 4.x process flow diagrams is performed using a single Group Processing node. Group processing in SAS Enterprise Miner 5.3 and later process flow diagrams is performed using two nodes, a Group Processing node and an End Group node. The End Group node defines the end of the portion of the process flow diagram that is intended for group processing. The End Group node in SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 also performs the model accumulation function that the Ensemble node performs in SAS Enterprise Miner 4.x.
A stratification model is an accumulation model that trains models over segments of the training data and then combines the scoring code from the individual models. Stratification models use a Group Processing node and a modeling node to create a separate model for each level of the stratification variable. The Ensemble node in SAS Enterprise Miner 4.x accumulates the multiple models built using stratified variables. As seen in this example, the End Group node performs the model accumulation task in SAS Enterprise Miner 5.3 and later:
Bagging and boosting models are accumulation models that resample the training data and create separate models for each sample. The predicted values (for models with interval targets) or the posterior probabilities (for models with class targets) are then aggregated to form the ensemble model. In SAS Enterprise Miner 4.x, the Ensemble node performs model accumulation for bagging and boosting models. In SAS Enterprise Miner 5.3 and later, the End Group node performs model accumulation for bagging and boosting models.
The following example shows a SAS Enterprise Miner 4.x process flow diagram that uses an Ensemble node to perform model accumulation for a boosting model. The example also shows a converted SAS Enterprise Miner 5.3 process flow diagram where the model accumulation and ensemble is performed by the End Group node.
The last example shows a SAS Enterprise Miner 4.x process flow diagram that uses Ensemble nodes both to combine models and to perform model accumulation for group processing. The converted SAS Enterprise Miner 5.3 and later process flow diagram uses an Ensemble node to perform model combination and an End Group node to perform model accumulation.
Interactive Grouping Node
When you convert a SAS Enterprise Miner 4.x process flow diagram that contains an Interactive Grouping node for use with SAS Enterprise Miner 5.3 and later, the converted diagram will contain one of two nodes, as follows:
If your SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 license does not include Credit Scoring for SAS Enterprise Miner, the Interactive Grouping node in your SAS Enterprise Miner 4.x process flow diagram is converted to an Interactive Binning node in the SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 process flow diagram.
If your SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 license includes Credit Scoring for SAS Enterprise Miner, the Interactive Grouping node in your SAS Enterprise Miner 4.x process flow diagram is converted to an Interactive Grouping node in the SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 process flow diagram.
Principal Components and DMNeural Nodes
In SAS Enterprise Miner 4.x, the Princomp/DMNeural node can be configured to perform either DMNeural network training or principal components analysis. SAS Enterprise Miner 5.3 and later uses two nodes to perform these functions. The SAS Enterprise Miner 5.3 and later DMNeural node performs DMNeural network training, and the SAS Enterprise Miner 5.3 and later Principal Components node performs principal component analysis.
SAS Enterprise Miner 4.x process flow diagrams that contain a Princomp/DMNeural node, when converted for use with SAS Enterprise Miner 5.3 and later, will contain either a DMNeural node or a Principal Components node, depending on how the SAS Enterprise Miner 4.x Princomp/DMNeural node was configured.
Use the General tab of the SAS Enterprise Miner 4.x Princomp/DMNeural node to specify whether to perform DMNeural network training or principal components analysis. DMNeural network training is performed when the Dmneural box is checked, and principal components analysis is performed when the Only do principal components analysis box is checked. Both boxes cannot be checked at the same time.
If the Princomp/DMNeural node in the SAS Enterprise Miner 4.x process flow diagram is configured to perform DMNeural network training, the converted SAS Enterprise Miner 5.3 and later process flow diagram will replace the Princomp/DMNeural node with a DMNeural node:
.
If the Princomp/DMNeural node in the SAS Enterprise Miner 4.x process flow diagram is configured to perform principal components analysis, the converted SAS Enterprise Miner 5.3 and later process flow diagram will replace the Princomp/DMNeural node with a Principal Components node:
Tree Node
The SAS Enterprise Miner 5.3 and later Decision Tree node uses an improved version of the splitting algorithm that was used in SAS Enterprise Miner 4.x. As a result, the improved Decision Tree algorithm in SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 might produce results that vary slightly from the results of the SAS Enterprise Miner 4.x Tree node.
Link Analysis Node
SAS Enterprise Miner 4.x uses the Link Analysis node to examine the linkage between effects in complex systems. SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 do not support Link Analysis. When you convert a SAS Enterprise Miner 4.x process flow diagram that contains a Link Analysis node, the resulting SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 process flow diagram displays a nonfunctional placeholder for the Link Analysis process. The placeholder Link Analysis node in SAS Enterprise Miner 5.3 and SAS Enterprise Miner 6.1 do not perform any analytical calculations. The placeholder node is able to display only the results that were generated by the Link Analysis node in the original SAS Enterprise Miner 4.x process flow diagram.
Reporter Node
The HTML report files that the SAS Enterprise Miner 4.x Reporter node creates are not transferred to converted SAS Enterprise Miner 5.3 and later projects. If your converted SAS Enterprise Miner 5.3 and later project contains process flow diagrams that use the Reporter node, you can run the process flow diagrams in SAS Enterprise Miner 5.3 or SAS Enterprise Miner 6.1 and save the Reporter node output as an LST or PDF file in the project's Reports folder.
Subdiagrams
The Subdiagram node in SAS Enterprise Miner 4.x is not supported in SAS Enterprise Miner 5.3 and later. If you convert a SAS Enterprise Miner 4.x process flow diagram that uses a Subdiagram node, the converted SAS Enterprise Miner 5.3 and later process flow diagram will not have a Subdiagram node. Instead, the converted SAS Enterprise Miner process flow diagram will accurately reproduce the detailed diagram substructure that was represented by the Subdiagram node in SAS Enterprise Miner 4.x. The subdiagram portion of the process flow diagram will be enclosed between Enter and Exit nodes.
Target Profiler
In SAS Enterprise Miner 4.x, you could use the Target Profiler to define or modify a target profile. Good target profiles produce optimal decisions that are based on a user-supplied decision matrix and output from a subsequent modeling procedure. In SAS Enterprise Miner 4.x you could access the Target Profiler through the Variables tab of any modeling node. In SAS Enterprise Miner 5.3 and later, you can only access Target Profiler information through an Input Data node or a Decision node. If you convert a SAS Enterprise Miner 4.x process flow diagram that has one or more modeling nodes that use the Target Profiler, the converted SAS Enterprise Miner 5.3 and later process flow diagram will add a Decision node before each modeling node that defined or modified the Target Profiler in SAS Enterprise Miner 4.x.
For example, consider the simple SAS Enterprise Miner 4.x process flow diagram here. The Target Profiler is used to specify prior probabilities for the target before modeling, as shown in the Prior tab of the Target Profiles for the GOOD_BAD window:
When the SAS Enterprise Miner 4.x project that contains this process flow diagram is converted for use with SAS Enterprise Miner 5.3, the converted diagram contains a Decisions node that precedes the Regression node. The Decisions node in the SAS Enterprise Miner 5.3 and later process flow diagram contains the prior probabilities that were specified for the target variable GOOD_BAD in the Target Profiler of the SAS Enterprise Miner 4.x Regression node.
Model Manager
All modeling nodes in SAS Enterprise Miner 4.x contain a utility called Model Manager. The Model Manager utility acts as a common framework that can be used to compare models and predictions, create assessment charts, and configure model reports.
In modeling nodes that aggregate models through stratification or looping, the SAS Enterprise Miner 4.x Model Manager allows you to view individual model results and fit statistics.
SAS Enterprise Miner 5.3 and later does not contain a Model Manager utility. If you convert this SAS Enterprise Miner 4.x process flow diagram that contains a modeling node that aggregates models as shown in this example, the modeling node in the converted SAS Enterprise Miner 5.3 and later process flow diagram retains only the model results from the last active model.This example shows the Model Manager window for a Tree node that is part of a group processing process flow diagram. When this process flow diagram is converted for use with SAS Enterprise Miner 5.3 and later, only the results from the last active model are transferred as SAS Enterprise Miner 4.x results.