DM: Re: forecasting

[Harm Smit]

Dear colleague,

Let me first introduce myself. My name is Harm Smit, my company is 
B3I. B3I
is a French software & systems house with data mining as one of its
activities. In this area, we are the French representatives of the 
Dutch
company SMR and distribute SMR's DataDetective data mining engine as 
well
as specific applications built upon this tool. DataDetective is based 
on
adaptive neural network techniques developed by SMR and has been 
applied in
a variety of areas - including database marketing, media planning, job
intermediation, criminology,  text mining, etc. - since about 4 years.
DataDetective supports associative ("fuzzy") searching and comprises 
an
interactive segmentation tool, named LookingGlass, allowing to 
discover,
visualize and analyze homogeneous subgroups of a data set.

Now, in answer to your question, DataDetective also includes a 
prediction
tool which perfectly fits your needs, assuming at least that you 
avail of
sufficient data from which the tool can learn. 
I think that further information should be exchanged off-line, 
otherwise
the "no commercial products advertising" netiquette would be 
compromised.

Best regards,

Harm SMIT
B3I
31 rue de l'Eglise
91430 VAUHALLAN
France
Tel. +33 1 69411863
Fax +33 1 60190981


----------
> De : Miguel Feldens <feldens@inf.ufrgs.br>
> A : datamine-l@nautilus-sys.com
> Objet : DM: forecasting
> Date : vendredi 15 août 1997 16:14
> 
> Hi!
> 
> I am looking for information on forecasting. The particular problem 
>I am
> starting to study, is a "shoe factory", a domain with lots of 
>factors and
> trends hard to be detected, which have strong influence on the kind 
>of
> shoes that will be produced. The kind of knowledge that would be
> interesting to discover in anything that can give good indications 
>about
> what kinds of shoes will be asked by the market, what the amounts, 
>and so
> on. The approach adopted by the company right now is to request
information
> from their main buyers, and trying to forecast production from such
> information, but that often results on bad predictions.
> 
> Has anyone done some work on a similar domain? Where can I find some
> references?
> 
> Thanks you all in advance!
> 
> Miguel Feldens
>