The first issue is: what does your Expert System do? What problem
can it solve, what situations do you see it being used in? What is
its "Domain"?

--
MikeC

Please reply to the group.

Can someone please help me on expert system
I am doing my theses about, building an expert system, using Visual Basic.
The problem is I am not a computer science student, I am an Industrial
engineer student.
I  have no background in programming although have been catching up with
both VB and Expert system.

My question is:

1. In building expert system (ES), I have found 2 books that actually
explains how to build it using basic (not VB). Problem is both are different
in method.
The other books dont seem to be clear enough on how to actually program the
ES.

-- one method is to create a matrix/table to assign values (of 1 or 0)to the
solution.
with solution being the coloumns and problems being the rows.
I created an excel spreadsheet for the rule, and get the user input values
for each parameters (row).
then ASSUMES that this is solution (coloumn) x.

Then another table will calculate using a formula D=D+R*V........(is this
familiar/correct formula to be used?)

when the result is + then the ASSUMPTION is correct and that particular
coloumn is the answer, if  -ve then ASSUME another coloumn (do-loop)
Problem is,  this method doesn't work, I kept getting the wrong answer.

2.  the other book creates somekind of database and compares the input to
database. Which is to me somewhat similiar to search method. (if x = x then
ok)
Although this method seems OK but then the system is merely comparing and
hence not quite an ES?
(the title of the book is something to do with ES though)

I am sorry for such long question,
any help will be greatly appreciated (GREATLY appreciated)
thank you in advance, at least for reading the whole question.

2/ you should know, which inferencial mechanism you want to use
(pseudoprobable, logical, standard, neural?)

3/ negative values cannot influence the final decision, unless you
explicitly say they can. i.e. you have to explicitely tell
    -A -> B(w)
where negative answer A leads to goal B with weight w

     A -> B(w)
goal B will be only influenced by positive answers, negative or zero answers
will leave goal B intact.

e.g.
question: Can animal fly?

     A -> B(w)
    animal can fly -> bird (3)

if the animal cannot fly you cannot say, it's not a bird unless you specify
    -A -> B(w)
    animal cannot fly -> bird(-3)

hth.
j.

Using VB is somewhat a "constraint" for me.

As for the Expert system, I need to do it for a case study of an unreliable
machines which requires the operator to be experienced. I will need to
"capture" his decision, such as what to do when what problems/parameter
occurs.

So far, I am aware of rule induction method, the if-then method and the one
I'm using, which is sort of matrix and value assigning system. (I may be
talking rubbish here but I was browsing through many books, so sorry for the
lack of understanding)

How would you approach this problem? and what book(s) did you use when
learning E.S?
(I found that most books about this subject is somewhat incomplete....and
relies heavily on prolog/natural languages--which is probably good, but not
for me)

Thank's again Mike.

1. Special cases where the ES throws up its hands, displays a message
   and beeps a lot. This is first pass.

2. Once the main ES is ready, you can attack these special cases again.

For the heart of the system which deals with the normal range of
inputs and outputs (data and control) of the machine, you might
be able to use decision tree structure. What you described sounds like
an alternate representation that's more search-based. It also lends
itself to some amount of fuzzy matching.

Using VB for this shouldn't be difficult.

It is too early to think in terms of data structures, in any language. It
is only after you get a thorough understanding (and appreciation) of the
nature of data you need to represent that the best DS to do it in can be
decided. You can always seek advice here about appropriate DS once you
understand your data and can explain it here.

Good luck.

--
MikeC

Please reply to the group.

Thank's again for replying

I'll try looking up the books you specified