BookmarkSubscribeRSS Feed
Bruno_F
Calcite | Level 5
Hi,

I try to use the HPBNET node in SAS EM in order to explain a target variable from input variables. By using this node, i would like to know the links between input variables and the effects on my target variable.
I tested several BN structures but only my target variable is a parent in my graph..
Can someone explain to me why my target is always a child ? How is it possible to make predictions if there is no parent variable for my target variable ?

Thanks !

Bruno
2 REPLIES 2
WendyCzika
SAS Employee

It depends on the structure of your network.  With a Parent-child or Markov blanket structure, the inputs can be parents of the target.  This paper has all the details and equations of how the predictions for a target are obtained from a Bayesian network: https://support.sas.com/resources/papers/proceedings17/SAS0474-2017.pdf

 

 

Bruno_F
Calcite | Level 5

Hello WendyCzika,

Thank you for your answer !

 

I built a Bayesian Network Model in order to explain a binary variable. Is it possible to compute a kind of "contribution" in order to know which input variable is the most contributive in my model to explain my target ?

 

Also, I picked up the Scoring code from my model but I don't understand the rules on it :

Can someone can explain to me what is compute in this rule (in red) because is not a probability..


else if _I28 = 1
then do;
  _target_score_HPBNC1{1}+(-1.538266175076);
  _target_score_HPBNC1{2}+(-0.241757117828);
end;

 

Best regards,

 

Bruno

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

Register now!

How to choose a machine learning algorithm

Use this tutorial as a handy guide to weigh the pros and cons of these commonly used machine learning algorithms.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 2 replies
  • 1026 views
  • 0 likes
  • 2 in conversation