mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sean Owen <sro...@gmail.com>
Subject Re: Persistent Data Model
Date Mon, 14 May 2012 11:54:05 GMT
Can you persist a DataModel? sure. The easiest thing is to read/write
a CSV file. Or put the data in a database. The existing
implementations already read such things into memory for you. I am not
sure what you mean about computing the DataModel as a separate
process. The DataModel exists and should be used by both online and
offline processes. No need for two.

Can you persist your recommendations? surely, however you like. Create
them, store them in a file or database. That's pretty up to you.


If your problem is such that you can afford to pre-compute many
recommendations (that is, you can afford to compute recommendations
you may not use, and, it's OK if recs are not up-to-the-second) then
your problem is a lot easier. Just run a process (unrelated to Tomcat)
that computes recs continuously in the background and publishes them
to some store, like a file system or database.  Hadoop is a fine fit
for this kind of use case here. While it may be overkill, its
drawbacks (batch-oriented-ness) would not be an issue.

Also look at using CachingRecommender to cache frequent recs, rather
than bother with all this. Maybe that's a solution.



On Mon, May 14, 2012 at 12:44 PM, Nikolaos Romanos Katsipoulakis
<popanik@gmail.com> wrote:
> Hi to everybody,
> I am currently developing a recommendation system, that relies on an
> Apache Tomcat Server (6.0.26) and is
> triggered by Web-Service Calls (JAX-WS).
> In order to improve the performance of the system, I have been asked to
> asynchronously call the recommender and save the recommendation
> data models in persistent storage (MySQL database), in an incremental
> manner. For example, I would have a ComputeRecommendations class that
> would create the data model, which would produce the recommendations.
> Can I save this class in persistent storage so that, at a later time, if
> I need to re-compute the recommendations, I would not need to create my
> data-model from the start?
>
> Thank you
>

Mime
View raw message