Mahout, mail # user - Kmeans cluster mapping to actual document IDs

Kmeans cluster mapping to actual document IDs

Hossein Kazemi 2012-04-11, 10:15

Hi,
I have clustered a set of documents using the Mahout's Kmeans
(map-reduce) I used Sparse Vectors due to the large size of my corpus.
In the book it says that the folder named ClusteredPoints contains the
mapping between the clustered documents and the document IDs. However,
all I can see is just a "1:0" , a feature-vector and a ClusterID. where
can I find the actual document names/ids ?
thx

Re: Kmeans cluster mapping to actual document IDs

Baoqiang Cao 2012-04-11, 16:20

My very limited experience is that

in seq2sparse step, you need use "-nv" option so that in clusterdump
output, you will see document ID.

Best,
Baoqiang

On Wed, Apr 11, 2012 at 5:15 AM, Hossein Kazemi <[EMAIL PROTECTED]> wrote:
> Hi,
> I have clustered a set of documents using the Mahout's Kmeans (map-reduce) I
> used Sparse Vectors due to the large size of my corpus. In the book it says
> that the folder named ClusteredPoints contains the mapping between the
> clustered documents and the document IDs. However, all I can see is just a
> "1:0" , a feature-vector and a ClusterID. where can I find the actual
> document names/ids ?
> thx
>