There are four extension
nodes that run JMP platforms and display results in SAS Enterprise
Miner.
-
JMP Boosted Tree —
runs the
Boosted Tree option of the JMP Partition
platform to produce a gradient boosting model. This node requires
JMP Pro.
-
JMP Bootstrap Forest —
runs the
Bootstrap Forest option of the JMP
Partition platform. Bootstrap forests are a variation of random forests,
a special type of bagging where predictors are selected at random
for each split. This node requires JMP Pro.
-
JMP Data Exploration —
opens a SAS Enterprise Miner data set and runs the JMP Graph Builder.
By default, it displays a graph of the distribution of each target.
But, you can interactively adjust the graphs to show, for example,
bivariate relationships between targets and predictors. Or, you can
create a new graph or analysis within JMP. For large data sets, you
can specify that the
JMP Data Exploration node
should use a sample of the input data set.
-
JMP Neural —
runs the JMP Neural platform, which produces a neural network model.
The node has a built-in gradient boosting option. Specify a positive
value for
Component Models to use boosting.
This node requires JMP Pro.
All nodes are compatible
with SAS Enterprise Miner data partitions, but only the modeling nodes
actually honor partitions. The
Data Exploration node
displays unpartitioned results by default. The
JMP Neural node
honors the SAS Enterprise Miner partition, but if you disable partitioning,
JMP will always perform
holdback or
K-fold validation,
determined by the user setting in
Fitting Options.