I took an online data-mining course that used this text as the introduction
to association rules
http://www.cs.waikato.ac.nz/~ml/weka/book.html<http://www.cs.waikato.ac.nz/%7Eml/weka/book.html>.
It didn't cover the specifics of the algoritms, but used WEKA. But it gives
you the idea. And you could even use the WEKA routines to process your
database.
Cindy Harper, Systems Librarian
Colgate University Libraries
charper_at_colgate.edu
315-228-7363
2010/7/1 Anders Söderbäck <Anders.Soderback_at_kb.se>
> As far as I am aware, the algorithms developed at Los Alamos and used in bX
> are available. I assume they can be found at the MESUR website (
> www.mesur.org), though I haven't read them myself. However, one of the
> developers for our LIBRIS union catalog (Henrik Lindström) read a few
> articles about user metrics before writing the algorithm used for our
> "related titles" application. Henrik mentioned he had seen the bX algorithm,
> but decided on some other way of solving the issue. (This is above my skills
> range too, but it is not rocket science. You just need to have good
> developers!)
>
> Our recommendation service is described in swedish at
> http://blog.libris.kb.se/librisutv/?p=731. Anyone interested can try
> running that page through google translate. Since we don't handle
> circulation in LIBRIS, the service is based on search logs. An exampel of a
> record with related items can be found here:
> http://libris.kb.se/bib/4740668
>
> The api is available but not as yet documented anywhere. However, you can
> try http://api.libris.kb.se/relrec/related?isbn=0-262-22029-6&format=xml
>
> We will have a discussion on releasing the data. Possibly it could be
> included in our linked data graph.
>
> As for bX, since SFX makes the usage data available through OAI-PMH,
> someone could harvest as many SFX-installations as possible (depending on
> how many libraries that would give access), run it through a recommendation
> algorithm (possibly the MESUR algorithm), and release the aggregated
> dataset. Maybe it could be called xB...
>
>
> best regards
> Anders Söderbäck
>
>
>
> -----Ursprungligt meddelande-----
> Från: Next generation catalogs for libraries genom Michele Newberry
> Skickat: on 2010-06-30 18:36
> Till: NGC4LIB_at_LISTSERV.ND.EDU
> Ämne: Re: [NGC4LIB] next "next-generation library catalog"
>
> > I certainly think that libraries could collaborate on the circ-based
> > recommendations. The advantage that bX has is that it is built on the
> > work done by Herbert and Johan at Los Alamos and I can attest that this
> > if far above my skills range. Does anyone know if their analysis
> > algorithms are available?
> >
> > - Michele
> >
>
Received on Thu Jul 01 2010 - 10:59:54 EDT