The HPBNET Procedure
Example 4.5 Bayesian Network-Augmented Naive Bayesian Network
This example also uses the Sampsio.Hmeq sample data set to learn a BAN structure, which is specified by STRUCTURE=NAIVE.
proc hpbnet data=sampsio.hmeq numbin=10 alpha=0.1 structure=Naive
missingint=IMPUTE missingnom=LEVEL;
target Bad;
input Reason Job Delinq Derog Ninq/level=NOM;
input Loan Mortdue Value Yoj Clage Clno Debtinc/level=INT;
output network=network;
run;
The following statements produce Figure 8, which shows the network structure. In the structure, Bad is a parent of Delinq, Derog, Job, Ninq, Clage, Clno, Loan and Mortdue. In addition, Delinq is the other parent of Derog, Job is a parent of both Mortdue and Clno.
proc print data=network noobs label;
var _parentnode_ _childnode_;
where _type_="STRUCTURE";
run;
Figure 8: BAN Network Structure
| Parent Node | Child Node |
|---|---|
| BAD | DELINQ |
| BAD | DEROG |
| DELINQ | DEROG |
| BAD | JOB |
| BAD | NINQ |
| BAD | CLAGE |
| BAD | CLNO |
| JOB | CLNO |
| BAD | LOAN |
| BAD | MORTDUE |
| JOB | MORTDUE |
Last updated: May 25, 2022