Re: Fuzziness: what am I missing?

Josef Feit (jfeit@dior.ics.muni.cz)
Sun, 5 Apr 1998 23:53:20 +0200 (MET DST)

"Michael D. Kersey" <mdkersey@hal-pc.org> writes:

>I don't know about others, but I see FL as an extremely useful technique
>for capturing ordinal information from language. Things that we
>previously talked about and intuitively "knew" can be captured in a
>mathematical model and included in a computer programme. E.g., "The
>robber was tall (short, fat, thin, dark-skinned, young, old, quick,
>slow, dim-witted, etc.)." FL provides techniques to put this into a
>database and do comparisons: e.g., "From our list of know felons, who
>best matches the above description." Incredible yet so incredibly easy!
>Not that there isn't debate about the details (especially here where
>we're mixing FL with decision theory).
>Keeping the faith,
>Michael D. Kersey

Can you please give any references and/or examples
of solving problems mentioned above?

I need to do some comparisons of entities described
in a similar way (skin diseases to be exact).

My solution uses scoring. Perhaps other approach could
be better?
Which person resembles a fat, old robber better:
young fat or old slim suspect?

Evaluating the importance of individual features is necessary,
and I cannot get it working right. (but there are problems
with information avilable, with different importance of
the features depending on the setting etc.)

Another problem is incomplete descriptions of diagnostic
units (= not only the description of a suspect, but the
felony database is incomplete).

Common fuzzy comparators seem to underestimate
or neglect secondary, less important signs.

I would appreciate any literary pointer to similar
problem solution.

J. Feit