How to use Mahout's model recommender in online experiments ?

[Alessandro Dias <al...@gmail.com>]

Hi,

I learned in this site below how to use ALS facorization algoritm to made
recommendations in Mahout Framework.

https://mahout.apache.org/users/recommender/intro-als-hadoop.html

From this:
- we inform a file with the rating (user, item, rating), in my case I have
 implicit ratings;
- then get the files of the two latent matrices generated; and
- finally we insert theses files in a recommender engine that generate a
file with the list of recomendations for each user.


I think that it is made for big e-commerce companies periodically. (the
model and recomendations is built periodically in an offline moments)



At my case, I'm going to do an online experiment of recommender. This model
recommender will be the control group.

I have a file with ratings of a set of old users and I will have a set of
new users on this online experiment. The old users will not participate
this experiment.

Theses new users will use the recommener system for 2 weeks in the online
experiment.



>> How to use ALSWRFactorizer recommender (non-hadoop) from Mahout in
online experiments ?

I'd like to build a model once and use it to the new users...

>> Will I have to run the algoritm (re-buid the model) in each
recomendation made during the online experiment ?

Thanks and Regards,

Alessandro Dias

Re: How to use Mahout's model recommender in online experiments ?

[Trevor Grant <tr...@gmail.com>]

Sorry for late response on this.

Might be worth checking out:
https://github.com/rawkintrevo/fsf17-twitter-recos

This is the corresponding talk. (relevant part starts at about 18:30)
https://youtu.be/h3j1JdtbhOI


Trevor Grant
Data Scientist
https://github.com/rawkintrevo
http://stackexchange.com/users/3002022/rawkintrevo
http://trevorgrant.org

*"Fortunate is he, who is able to know the causes of things."  -Virgil*


On Thu, May 25, 2017 at 9:51 AM, Alessandro Dias <al...@gmail.com>
wrote:

> Hi,
>
> I learned in this site below how to use ALS facorization algoritm to made
> recommendations in Mahout Framework.
>
> https://mahout.apache.org/users/recommender/intro-als-hadoop.html
>
> From this:
> - we inform a file with the rating (user, item, rating), in my case I have
>  implicit ratings;
> - then get the files of the two latent matrices generated; and
> - finally we insert theses files in a recommender engine that generate a
> file with the list of recomendations for each user.
>
>
> I think that it is made for big e-commerce companies periodically. (the
> model and recomendations is built periodically in an offline moments)
>
>
>
> At my case, I'm going to do an online experiment of recommender. This model
> recommender will be the control group.
>
> I have a file with ratings of a set of old users and I will have a set of
> new users on this online experiment. The old users will not participate
> this experiment.
>
> Theses new users will use the recommener system for 2 weeks in the online
> experiment.
>
>
>
> >> How to use ALSWRFactorizer recommender (non-hadoop) from Mahout in
> online experiments ?
>
> I'd like to build a model once and use it to the new users...
>
> >> Will I have to run the algoritm (re-buid the model) in each
> recomendation made during the online experiment ?
>
> Thanks and Regards,
>
> Alessandro Dias
>