The JMP Extension Node
package requires the following:
-
SAS Enterprise Miner, version 6.2
or later. The modeling nodes are single machine tools. Currently,
they do not work in a client/server, multi-machine deployment of SAS
Enterprise Miner.
-
JMP Pro, Version 9.0.3. There is
no error checking to make sure you have JMP Pro 9 on your system.
Errors might occur if you use a standard version of JMP, rather than
JMP Pro.
When using the JMP Extension
Node package, you should be aware of the following issues:
-
-
The modeling nodes, JMP Boosted
Tree, JMP Bootstrap Forest, and JMP Neural, require JMP Pro. They
fail with error messages when standard JMP is used.
-
The nodes work in a SAS Enterprise
Miner client/server configuration, provided that JMP Pro, the SAS
Server, and the SAS Enterprise Miner client are all on the same machine.
This includes SAS Enterprise Miner desktop, SAS Enterprise Miner classroom,
SAS Enterprise Miner workstation, or a three-tier, single-machine
install of SAS Enterprise Miner. The modeling nodes do not currently
run in a multi-machine deployment of SAS Enterprise Miner.
-
The modeling nodes accept a single
target variable only. If there are more targets, an error condition
is triggered.
-
-
SAS syntax errors can occur if
the input data set has many predictors, predictors with long names,
or both. If you see syntax errors along with a
Truncated
Record warning in the SAS log, set a high value for the
LRECL option in your project start-up code. Here is an example:
OPTIONS
LRECL=5000;
.
-
SAS syntax errors can occur with
long variable names (approximately 25 characters or more) and string
data values that have embedded quotation marks, such as "isn't". If
you encounter syntax errors, try using shorter variable names, and
transform any input strings that have embedded quotation marks or
apostrophes.
-
-
Under certain conditions, there
can be discrepancies between the results reported by JMP and those
reported by SAS Enterprise Miner. To avoid these discrepancies, set
the
Number of Terms property to its default
value of
1 for Bootstrap Forest models, impute
missing data values, or both. Specifically, if you have missing predictor
values, use the default value for the
Number of Terms property.
If you need to change the
Number of Terms property,
then impute missing predictor values before you run the node.