New GeNIe user

The front end.
Post Reply
Elise2992
Posts: 3
Joined: Mon Feb 06, 2023 9:26 am

New GeNIe user

Post by Elise2992 »

Hello,

For my master's thesis in forensic science I am working on a BN and have some questions since I am new to the theory behind the networks and the GeNIe program.

1. When making a BN using the program and wanting to do a validation (K-fold, leave-one-out) afterwards do I need to divide my dataset in a training and test set before making a BN?

2. I want to do a sensitivity analysis on my BN and do not know how to pick the right target. I thought using my class variable, that I want to predict, would be good but then I only get a color change to red for this variable. I think I do not understand the complete concept of this and the manual I use does not seem to clear things up for me.

Thank you for your help!
marek [BayesFusion]
Site Admin
Posts: 430
Joined: Tue Dec 11, 2007 4:24 pm

Re: New GeNIe user

Post by marek [BayesFusion] »

Hi Elise2992,

Welcome to GeNIe/BN community! Let me try to answer your questions quickly:

> 1. When making a BN using the program and wanting to do a validation
> (K-fold, leave-one-out) afterwards do I need to divide my dataset in
> a training and test set before making a BN?

You can but you don't have to. If you do divide the data set into two, please learn your model from the training data set and perform validation using "Test only," which will precisely just test your model on the test data set. If you use "K-fold crossvalidation" (or its extreme case "Leave one out"), GeNIe will split the data into training and test subsets automatically.

> 2. I want to do a sensitivity analysis on my BN and do not know how to
> pick the right target. I thought using my class variable, that I want to predict,
> would be good but then I only get a color change to red for this variable. I think
> I do not understand the complete concept of this and the manual I use does not
> seem to clear things up for me.

The choice of the class variable is correct. At least, this is what I would do. If you have no evidence or no evidence in the descendants of your class variable, none of the parameters in the descendants will turn red (or pink). The ancestors should show colors, though. Once you enter evidence (these are going to be your test cases), you should see colors among the descendants. If you don't, your parameters may be not set correctly. If you cannot find the problem, please post your model or its subset and I will be happy to comment on it.

I hope this helps,

Marek
Post Reply