[ 
https://issues.apache.org/jira/browse/MAHOUT-954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13192125#comment-13192125
 ] 

Ikumasa Mukai edited comment on MAHOUT-954 at 1/25/12 7:39 AM:
---------------------------------------------------------------

The output of RegressionResultAnalyzer is changed like this.
{noformat}
=======================================================
Summary
-------------------------------------------------------
Correlation coefficient                 :     0.4257
Mean absolute error                     :     0.9395
Root mean squared error                 :     1.2048
Predictable Instances                   :        717
Unpredictable Instances                 :          0
Total Regressed Instances               :        717
{noformat}
2 items, "Predictable Instances" and "Unpredictable Instances" are added.

For getting them, my patch contains the modification for classify() on 
DecisionForest.java.
With the trunk version of this method, we can get just a last tree, not all 
trees.
                
      was (Author: ikumasa mukai):
    The output of RegressionResultAnalyzer is changed like this.
{noformat}
=======================================================
Summary
-------------------------------------------------------
Correlation coefficient                 :     0.4257
Mean absolute error                     :     0.9395
Root mean squared error                 :     1.2048
Predictable Instances                   :        717
Unpredictable Instances                 :          0
Total Regressed Instances               :        717
{noformat}
2 items, "Predictable Instances" and "Unpredictable Instances" are added.

                  
> "Unpredictable" have to be represented by NaN on DF.
> ----------------------------------------------------
>
>                 Key: MAHOUT-954
>                 URL: https://issues.apache.org/jira/browse/MAHOUT-954
>             Project: Mahout
>          Issue Type: Bug
>          Components: Classification
>            Reporter: Ikumasa Mukai
>              Labels: DecisionForest, Regression
>         Attachments: MAHOUT-954.patch
>
>
> On our RandomForest src, we uses "-1" for representing "Unpredictable",
> but for regression problem, there is a case which outputs "-1" for the 
> correct result.
> For resolving this problem, we should change the unpredictable flag to "NaN" 
> from "-1".

--
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

        

Reply via email to