BNET Procedure

Example 5.2 Tree-Augmented Naive Bayesian Network

This example also uses the Iris data set that is available in the Sashelp library. In the following statements, STRUCTURE=TAN results in a tree-augmented Bayesian network:

data mylib.iris;
   set sashelp.iris;
run;
 proc bnet data=mylib.iris numbin=3 structure=TAN
             prescreening=0 varselect=0
             outnetwork=mylib.network;
     target Species;
     input PetalWidth PetalLength SepalLength SepalWidth/level=interval;
 run;

The following statements produce Output 5.2.1, which shows the network structure. In the structure, Species is a parent of PetalWidth, PetalLength, SepalLength, and SepalWidth. In addition, PetalWidth is a parent PetalLength, SepalWidth is a parent of SepalLength, and PetalWidth is a parent of SepalWidth.

 proc print data=mylib.network noobs label;
     var _parentnode_ _childnode_;
     where _type_="STRUCTURE";
 run;

Output 5.2.1: TAN Network Structure

Parent NodeChild Node
SpeciesPetalLength
PetalWidthPetalLength
SpeciesPetalWidth
SpeciesSepalLength
PetalLengthSepalLength
SpeciesSepalWidth
SepalLengthSepalWidth


Last updated: August 06, 2026