DM: Information on Diagnosis

[Shlomo Urbach]

Hello all,Some time ago I sent the following request to this list.
Some reponders requested that I publish a compilation of replies.
So here it is (I hope the attachments work well):
        first attachment is the original request.
        rest of attachments are replies.
If anyone wishes to add or comment, please email or post.

Shlomo Urbach
Neptune Software


Message-ID: <371586E5.7DF8B855@neptune.co.il>
Date: Thu, 15 Apr 1999 09:27:49 +0300
From: Shlomo Urbach <shlomo@neptune.co.il>
Organization: Neptune Software
MIME-Version: 1.0
To: datamine-l <datamine-l@nautilus-sys.com>Subject: DM: information 
on diagnosis

Hello,

Is anyone out there using DM (or Multivariate Regression or whatever)
in combination with some domain knowledge for diagnostic purposes ?

I'm working on extrating "probable cause" from large (100 - 100000 
records)datasets.
The target is to understand which of the many (~100) variables (X's) 
affect the outcome of a (single known) target variable (Y).
The algorithm is used to diagnose behavior of a large system, and
tell the user what causes poor behavior.
Some limited domain knowledge is available. It may be formulated as
hierarchies and loose cause/effect relationships between the various
X's.
NOTE: Y and most X's are numeric. Some X's are categorical.I'm 
interested in any papers, books, algorithms, etc. about similar
experiences.
Especially:
        1. Which algorithms used to detect the "interesting" (cause) 
X's ?
        2. How domain knowledge is integrated in the algorithms?
        3. How are results (which are always somewhat uncertain) 
shown to the novice user, who only wants to know the cause, and 
not any DM mumbo-jumbo ?

Thanks,

Shlomo Urbach
Neptune Software
shlomo@neptune.co.ilFrom: Donal Lyons <dlyons@stats.tcd.ie>
To:   <shlomo@neptune.co.il>
Subject: Re: information on diagnosis
Date: Thu, 15 Apr 1999 08:01:21 +0100

>Is anyone out there using DM (or 
Multivariate Regression or whatever)
>in combination with some domain knowledge for diagnostic
purposes ?

Shlomo,I don't have an answer to this but have a problem with some 
similarities and would be interested in any references you
 come up with.

I'm working with a daily set of all usages of pre-paid tickets on 
Dublin
buses - this is now 17 months old and 2.5Gb.  Domain knowledge of the
geography of stages is essential in understanding passenger movements 
but so far I've found no systematic way of building this in - we make 
various ad hoc arrangements.

Regards,