[
https://issues.apache.org/jira/browse/MAHOUT-1618?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14170146#comment-14170146
]
Pat Ferrel commented on MAHOUT-1618:
------------------------------------
https://guide.finderbots.com
This is an integration inside a video guide written in RoR using MongoDB. The
models created by spark-itemsimilarity are put into the catalog in MongoDB and
Solr indexes those models. The language used to put the models in Mongo could
be Python (though it was actually Java in this case).
I do not suggest you use this as template but you get the idea. Your idea is
fine though. A single language tutorial is better in some ways because it is
easier to read and the integration is simpler.
Solr is the central part so I'd work from there out. Solr indexes the model
created by spark-itemsimilarity. You then find user history and use it as a
query on the model. So using your template you'd have to build models with
Mahout and index them with Solr. Then in perhaps a second application you'd
identify a user to get their history, which is the query or allow people to
hand enter item ids corresponding to hypothetical history.
In this case you have:
1) a batch/background application that runs spark-itemsimilarity *and* indexes
the data with Solr. This could be a Scala program if you want or might even be
a bash script.
2) an application that builds the query from user history or allows you to
enter hypothetical user history. It would be nice if this was interactive to
allow the user to pick a user by ID or enter a list of item IDs as hypothetical
history and get recs.
These two apps would illustrate all integration points in a minimal manner. If
you want to be more ambitious go for it. You could build both together using
Play2 to create a web app in Scala that has a button for "build and index
models" and an input field for entering user history or user ID with another
button for "make recommendations".
> Cooccurrence Recommender example and documentation
> ---------------------------------------------------
>
> Key: MAHOUT-1618
> URL: https://issues.apache.org/jira/browse/MAHOUT-1618
> Project: Mahout
> Issue Type: Documentation
> Components: Examples
> Affects Versions: collections-1.0
> Reporter: Thejas Prasad
> Priority: Trivial
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
