Expert Systems?

[Alex Rice]

On Tuesday, October 8, 2002, at 04:18  PM, David Vaughan wrote:

> Can you reduce those formulae ultimately to a single score (which may 
> be a weighted combination of sub-scores)? If so, it sounds to me that 
> a GA could do what you want and it is not difficult to implement in 
> Rev. GAs are especially good at munging multiple issues, just so long 
> as, eventually, you can single-score a solution. Mixing score 
> weightings to achieve that is something you can do with user controls 
> rather than embedding, if you are unsure of what the weightings should 
> be.

Genetic algorithms are mainly useful for function optimization 
problems, I think.

This problem is really perfect for an expert system because it's, well, 
an expert system. :-) A bunch of heuristics, rules-of-thumb, which when 
you apply them to set of data set it outputs some recommendations 
recommendations. Stuff like  "If you employ spatula repair persons, 
then you get a spatula repair bay. If the facility is in a snowy clime, 
you get an indoor spatula repair bay with kitchenette. If not in a 
snowy clime, you get an outdoor spatula repair bay. You get a small, 
medium or large spatula repair bay if .... etc."

Alex Rice, Software Developer
Architectural Research Consultants, Inc.
alrice at swcp.com
alex_rice at arc.to