[
https://issues.apache.org/jira/browse/MAHOUT-825?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13121955#comment-13121955
]
Paritosh Ranjan edited comment on MAHOUT-825 at 10/6/11 1:54 PM:
-----------------------------------------------------------------
I have attached two versions of clustered points ( clusters ). This is real
data which my application was clustering. Opening this file in Notepad++,
EditPlus etc. will show you the results in better format. I have also printed
Vectors, but you can only see the Strings in each line. The functionality is,
that, similar looking Strings should be clustered together.
One file ( Clustering Remote Points - Two Big, Useless Clusters ) contains two
big clusters ( Scroll down to see the second cluster ). This file is created
using cluster all points approach.
Second File ( Not Clustering Remote Points - Two Meaningful Clusters ) contains
points clustered only within t1.
I hope you will agree that the second file with two small clusters makes sense
in this case.
Now I have used a flag to cluster data Strictly (instead of clusterFilter) . I
have created a patch and attached ( canopy-strict-clustering-flag ). It also
has test cases which demonstrate how remotely present points can be added or
declined from a canopy (cluster).
The cluster files that I have attached, are, the exact use case, for which, I
fixed this issue in Mahout. I encountered this problem, and I was clueless of
what is happening inside. I kept checking my vector creation and distance
measures, but the problem was not there. So, I think this patch can help
others. Also, the default option is false, so, it can be used as per user's
requirement.
was (Author: paritoshranjan):
I have attached two versions of clustered points ( clusters ). This is real
data which my application was clustering. Opening this file in Notepad++,
EditPlus etc. will show you the results in better format. I have also printed
Vectors, but you can only see the Strings in each line. The functionality is,
that, similar looking Strings should be clustered together.
One file ( Clustering Remote Points - Two Big, Useless Clusters ) contains two
big clusters ( Scroll down to see the second cluster ). This file is created
using cluster all points approach.
Second File ( Not Clustering Remote Points - Two Meaningful Clusters ) contains
points clustered only withing t1.
I hope you will agree that the second file with two small clusters makes sense
in this case.
Now I have used a flag to cluster data Strictly (instead of clusterFilter) . I
have created a patch and attached ( canopy-strict-clustering-flag ). It also
has test cases which demonstrate how remotely present points can be added or
declined from a canopy (cluster).
The cluster files that I have attached, are, the exact use case, for which, I
fixed this issue in Mahout. I encountered this problem, and I was clueless of
what is happening inside. I kept checking my vector creation and distance
measures, but the problem was not there. So, I think this patch can help
others. Also, the default option is false, so, it can be used as per user's
requirement.
> Canopies grouping records outside t1
> ------------------------------------
>
> Key: MAHOUT-825
> URL: https://issues.apache.org/jira/browse/MAHOUT-825
> Project: Mahout
> Issue Type: Bug
> Components: Clustering
> Affects Versions: 0.6
> Environment: windows, linux
> Reporter: Paritosh Ranjan
> Labels: features, newbie, patch
> Fix For: 0.6
>
> Attachments: Clustering Remote Points - Two Big, Useless
> Clusters.txt, Not Clustering Remote Points - Two Meaningful Clusters.txt,
> canopy-clusterFilter-t1, canopy-outside-t1-points-patch-1,
> canopy-strict-clustering-flag
>
>
> While finding closest canopy, there is no check to ensure that it returns
> canopies which are within distance t1 from the point. This results in
> incorrect result i.e. Points outside t1 are grouped in canopies.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
