[
https://issues.apache.org/jira/browse/MAHOUT-305?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12838044#action_12838044
]
Tamas Jambor commented on MAHOUT-305:
-------------------------------------
I usually pick random N % data for each user as Ankur suggested. This would
ensure that the recommender is not biased, and it doesn't really matter that
non-relevant items are in this subset, since they needed to be ranked lower
anyway.
I think the way Sean implemented it is also pretty good, taking the top-n
relevant items and evaluating on that data, but you have to build a new model
for each user, which makes it impossible to use it on a big data set,
especially with SVD.
I agree that the other issue is how to deal with non-rated item. I personally
just rank items that has known ratings, so that the relevance judgement is
always known. but I've been thinking to changing it to count unrated items as
non-relevant. I think there are pros and cons either way.
> Combine both cooccurrence-based CF M/R jobs
> -------------------------------------------
>
> Key: MAHOUT-305
> URL: https://issues.apache.org/jira/browse/MAHOUT-305
> Project: Mahout
> Issue Type: Improvement
> Components: Collaborative Filtering
> Affects Versions: 0.2
> Reporter: Sean Owen
> Assignee: Ankur
> Priority: Minor
>
> We have two different but essentially identical MapReduce jobs to make
> recommendations based on item co-occurrence:
> org.apache.mahout.cf.taste.hadoop.{item,cooccurrence}. They ought to be
> merged. Not sure exactly how to approach that but noting this in JIRA, per
> Ankur.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
