Re: Counting links between to records on the same table (Clustering?)

From: Dale Walker <dale_at_sorted.org>
Date: Mon, 27 Jun 2005 17:23:53 +0100
Message-ID: <so70c1dceqt0uce6v9kmtpno1sficvgbpv@4ax.com>

On Mon, 27 Jun 2005 15:34:58 GMT, paul c <toledobythesea_at_oohay.ac> wrote:

>Dale Walker wrote:
>> I'm trying to develop a little method that groups DJs together into
>> similar tastes in music so that people can work out if they might like
>> a certain night out.

.
.
.

>i suppose the canonical approach is to say that the relation is between
>events and dj's, not among dj's, so a composite key could be event id
>and dj id where the implicit total is always 1! or 1 factorial if you
>prefer. undoubtedly somebody will suggest constraints on the grounds
>that this key is not mapped well by conventional products but this is a
>theory group so that doesn't matter, haha.

Actually, I'm trying to divorce the information away from the events themselves as they too keep flitting and flip-flopping between styles and genres. They also have a habit of changing names, being the second production in a two roomed event and numerous other things that could upset the results due to me not recording all the data in the beginning.

>so my cut at it is that this is a query question, not a design question
>and i'll leave the sql to people who like that stuff.

I hope it is just a simple query problem but the whole impetus behind this query is to start getting into using AI techniques in databases to collate data. Thought it was a good idea to start small and work my way up.

I come from a coding background and find that it's a often tough job to reprogram my brain to deal with problems in terms of queries.

>i've given you perhaps no answer at all. it's just that it makes me
>wonder - i don't know about others but i often find myself wanting to
>impose some kind of geometry on such designs even though the rm doesn't
>intend that. personally, i'd like to see a product that lets me say two
>unpopulated 'views' or relations for that matter, are somehow equivalent
>without a bunch of obscure constraint coding. i don't know if such a
>concept has a name, perhaps co-dependent relations where the db asserts
>that you can't have one without the other or if it's all part of the
>user or psychological aspect of the rm or if it's not a concept at all
>and just logically ridiculous.

It's probably just a query question and you keep leaving it to the sql people to sort out ;)

---
Dale Walker
London Techno Events
dale_at_sorted,org
http://london.sorted.org

Received on Mon Jun 27 2005 - 11:23:53 CDT