[jira] [Created] (MAHOUT-1621) k-fold cross-validation in MapReduce Random Forest example?

["Tawfiq Hasanin (JIRA)" <ji...@apache.org>]

Tawfiq Hasanin created MAHOUT-1621:
--------------------------------------

             Summary: k-fold cross-validation in MapReduce Random Forest example?
                 Key: MAHOUT-1621
                 URL: https://issues.apache.org/jira/browse/MAHOUT-1621
             Project: Mahout
          Issue Type: Question
          Components: Examples
         Environment: Ubuntu Linux 14.04
            Reporter: Tawfiq Hasanin
             Fix For: 1.0


My goal is to modify MapReduce Random Forest example by combining BuildForest.java and TestForest.java into a new class called RandomForest.java

The main point is to input one data file which is going to be used in training and testing; with k-fold cross-validation. 

I have a big data with hight diminutional features and small amount of instances. 

Seems to be a frustrating dead-end. is this process achievable? Or is it against MapReduce nature? 

Thanks..



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Re: [jira] [Created] (MAHOUT-1621) k-fold cross-validation in MapReduce Random Forest example?

[Ted Dunning <te...@gmail.com>]

On Tue, Oct 14, 2014 at 10:35 AM, Tawfiq Hasanin (JIRA) <ji...@apache.org>
wrote:

> I have a big data with hight diminutional features and small amount of
> instances.
>
> Seems to be a frustrating dead-end. is this process achievable? Or is it
> against MapReduce nature?
>

Sounds a lot against map-reduce nature.

How is it your data is big if you only have a small number of instances?