Mailing-List: contact dev-help@mahout.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@mahout.apache.org
Received-SPF: pass (nike.apache.org: domain of vivekkhanna@hotmail.com
 designates 65.55.34.16 as permitted sender)
Message-ID: <COL124-W6E38FA7CD583D4160DE65ABC60@phx.gbl>
Content-Type: multipart/alternative;
	boundary="_85566ad4-ae9a-4b49-b221-00578af63cbb_"
From: Vivek Khanna <vivekkhanna@hotmail.com>
To: <dev@mahout.apache.org>
Subject: RE: User/Items Reco Engine clustering
Date: Thu, 24 Jun 2010 09:14:57 -0400
Importance: Normal
In-Reply-To: <AANLkTikxX8EQVB4hQ6dwoXbA_gUiD62BrWDQYygQv23_@mail.gmail.com>
References: 
 <AANLkTim5MKPUMgvLLOzPBEIbt1vSlnnI3djbjZnATHg4@mail.gmail.com>,<COL124-W46AA7FC6461144E98EE46BABC40@phx.gbl>,<AANLkTin6pmNID51CyZgTLm8PThJm00OWO0QpuMvDL6lx@mail.gmail.com>,<AANLkTikxX8EQVB4hQ6dwoXbA_gUiD62BrWDQYygQv23_@mail.gmail.com>
MIME-Version: 1.0

--_85566ad4-ae9a-4b49-b221-00578af63cbb_
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable


Another way to look at the problem is to consider user purchases/actions as=
 features describing a user in a vector space. Then the problem is reduced =
to finding users similar to each other based on this feature set.

Clustering would be overly complex in my humble opinion.

I agree with Sean that the Lucene based construction as you describe it Jay=
=2C is item-based and not user-based.=20

Hope this helps.

> Date: Wed=2C 23 Jun 2010 08:59:03 +0100
> Subject: Re: User/Items Reco Engine clustering
> From: srowen@gmail.com
> To: dev@mahout.apache.org
>=20
> To me you're just describing user-based recommendation. You find a
> neighborhood of similar users=2C then examine their items=2C and recommen=
d
> from those by taking a weighted average of the neighborhood's
> preferences.
>=20
> Your Lucene-based construction then sounds like item-based
> recommendation. Find items similar to what the user prefers and
> recommend based on a weighted average=2C again.
>=20
> Do I have that right?
>=20
> And then=2C do you need a Hadoop-based implementation using SequenceFiles=
?
> What kind of data size are you looking at?
>=20
> On Wed=2C Jun 23=2C 2010 at 12:49 AM=2C Jay Sellers <jaysellers@gmail.com=
> wrote:
> > Thanks Vivek=2C
> > We do not have predefined clusters/groups. We expect the groups to muta=
te as
> > more history (data) is accumulated.  A simple use case is as follows:
> > John has viewed a pair of jeans=2C a cowboy hat=2C a red shirt and a pa=
ir of
> > boots.
> > Scott has viewed a pair of jeans=2C a cowboy hat=2C a red shirt and a p=
ocket
> > watch.
> > Larry has viewed a pair of jeans=2C a cowboy hat and a red shirt.
> >
> > When we send Larry and his items into our reco engine=2C we would expec=
t a
> > pair of boots and a pocket watch to be recommended.  We'd expect this
> > because we've determined that John and Scott are 'like' Larry and thus =
are
> > in the same cluster.
> >
> > Again=2C we fully expect the cluster members to change=2C as user/item =
data
> > accumulates.
> >
> > On Tue=2C Jun 22=2C 2010 at 4:37 PM=2C Vivek Khanna <vivekkhanna@hotmai=
l.com>wrote:
> >
> >>
> >> Hi=2C
> >>
> >>
> >>
> >> For your clustering/grouping=2C what is your expectation? Do you have
> >> pre-defined clusters/groups that you want to cluster the items within =
those=2C
> >> or do you envision a system where clusters/groups will change and evol=
ve as
> >> the data changes?
> >>
> >>
> >>
> >> In each case=2C it seems you are looking for unsupervised approaches. =
Is that
> >> correct?
> >>
> >>
> >>
> >> I am new to this email list=2C so pardon my ignorance=2C but from what=
 work I
> >> have done in the past with IR=2C ML (clustering=2C More like this=2C
> >> categorization=2C topic detection etc.)=2C my advice to you is to iden=
tify your
> >> requirements=2C use cases and page flow interactions as the first step=
. :)
> >>
> >>
> >>
> >> Hope this helps!
> >>
> >> Vivek.
> >>
> >> > Date: Tue=2C 22 Jun 2010 15:50:18 -0700
> >> > Subject: User/Items Reco Engine clustering
> >> > From: jaysellers@gmail.com
> >> > To: dev@mahout.apache.org
> >> >
> >> > I'm looking to enhance a product recommendation engine. It currently
> >> works
> >> > with all data as a whole. I want to introduce clustering/grouping. I=
ts
> >> > model based and the relationship is the common User-Items relationsh=
ip.
> >> > Originally I was thinking of using a Canopy / kmeans cluster. And th=
en
> >> > determine which cluster a user is in and execute Item Similarity aga=
inst
> >> > only that cluster of items. However I can't figure out how to build =
a
> >> > SequenceFile using vectors with the User/Items relationship. I don't=
 know
> >> > which data points to feed the vector. So I scratched that idea and t=
urned
> >> > my attention to Lucene=2C thinking that this is really a document is=
sue.
> >> Where
> >> > users are documents and the items are the content. I should be able =
to
> >> ask
> >> > Lucene=2C give me documents that look like this "productId3 productI=
d9056
> >> > productId234".
> >> >
> >> > I'm looking for any and all feedback from those experienced in the
> >> > recommendation world=2C specifically with the grouping of users and =
items.
> >> >
> >> > Thanks=2C
> >> > -Jay
> >>
> >> _________________________________________________________________
> >> The New Busy is not the old busy. Search=2C chat and e-mail from your =
inbox.
> >>
> >> http://www.windowslive.com/campaign/thenewbusy?ocid=3DPID28326::T:WLMT=
AGL:ON:WL:en-US:WM_HMP:042010_3
> >
 		 	   		 =20
_________________________________________________________________
The New Busy is not the too busy. Combine all your e-mail accounts with Hot=
mail.
http://www.windowslive.com/campaign/thenewbusy?tile=3Dmultiaccount&ocid=3DP=
ID28326::T:WLMTAGL:ON:WL:en-US:WM_HMP:042010_4=

--_85566ad4-ae9a-4b49-b221-00578af63cbb_--