To be completely honest I am an extreme novice when it comes to Bayesian Belief Networks. I just came across the subject the day before yesterday, via the following document.

http://freelock.com/files/KE.pdf

Here's my task:

Create an "expert system" to assist help desk technicians with problem resolution.

This system should be accessible via a web interface.

Hypothetically speaking, it would be nice to cross an open source based belief network with an open source help desk ticketing system. If my assumptions are correct, over time the accuracy would increase.

========

It's safe to say that I'm pretty confused about the proper direction to go. MSBNx is only free for non-commercial user, so it's out of the question. However, everything I'm reading in the document above mirrors what I need: A Windows 98 style troubleshooter

Deviations from a Windows 98 troubleshooter:
Need web server based application
Ideally would be unix based
Open source
Some place for me to ask questions.. :-)

========

I'm not sure if OpenBayes would be a good choice. Please let me know your thoughts on the matter.

Thanks for your time,
Joshua

Hi Joshua,

>> Hypothetically speaking, it would be nice to cross an open source based belief network with an open source help
>> desk ticketing system. If my assumptions are correct, over time the accuracy would increase.

yep, that should work. Provided that a correct learning engine is used and that all of your data is used in the learning process. But don't take that as a radical answer. The answer to your question strongly depends on the BN you're planning to use and the data available.

As for your choice on the software to use, I can only give you a list of available libraries and software, but you must be the one to make the choice. Just a short list :
- BNT for MatLAB (Matlab...yack)
- PNL for C++ (very difficult to understand, few documentation, no support, but very very efficient)
- JavaBayes (didn't have the chance to use this one, rumors say that it's still in debugging process, as most of the bayesian software)
- ....

and some commercial software that come with an API
- Netica
- BayesiaLab
- MSBNX
- ...

I think that OpenBayes meets your demands (unix, opensource, ...) but you should keep in mind that it is still under development and results should be double checked at the beginning using another software. (for example I'm working on correcting a bug in the inference engine for the moment)

However, if you're interested in opensource software and would like to use OpenBayes, I can help you understand the code and make your life easier when it comes to implementing stuff that do not exist yet.

To be honest, I'm searching for people that are ready to use OpenBayes and improve it. My idea of opensource is software that do not belong to anybody, including the main programmer. In that sense, I'm hoping that one day this project will be mature enough to continue it's way alone, without my help (or maybe a little).

Think about it and keep in touch. Even if you choose another platform, please let us know the reasons that made you choose it. It is very interesting feedback for us.

Thank you for your interest
Kosta