Question about choice of a recommender

["Pasmanik, Paul" <Pa...@danteinc.com>]

Hi, I am trying to figure out what recommender or a combination of other techniques to use.
I have a system where a user is presented with a list of pre-defined (max 50 products) when he/she is viewing a video clip from a finite (not huge) set.    The relationship between video clip and what products to show is statically defined and can change from time to time.   We collect clicks and resulting buys for these products.   What we are trying to do is to come up with the recommender engine that would display top 4-5 products that a user would most likely click on and possibly buy (limited to the pre-defined list of up to 50 or so products for that particular video clip).
We can think of a set of these pre-defined products as a category, but a product can appear in multiple categories.

What would be the best approach?

Thanks.

________________________________
The information contained in this electronic transmission is intended only for the use of the recipient and may be confidential and privileged. Unauthorized use, disclosure, or reproduction is strictly prohibited and may be unlawful. If you have received this electronic transmission in error, please notify the sender immediately.

Re: Question about choice of a recommender

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

How much data are you going to be collecting?  How many users and how many
presentations per user?

Are you saying that the product for each video are completely fixed?  Does
the same product appear for more than one video?

Do users interact with products outside of the narrow confines that you
have described?

If you don't really have social information and the products are very
fixed, then simply learning to rank using Thompson sampling might be a good
idea.

http://tdunning.blogspot.com/2013/04/learning-to-rank-in-very-bayesian-way.html

If you do have enough information and it crosses users, then it might work
to build real recommenders.  If you are object space is very small and you
have user demographics, you might consider building a targeting model per
item.

You might also consider breaking open your constraints a bit.




On Mon, Dec 15, 2014 at 1:40 PM, Pasmanik, Paul <Pa...@danteinc.com>
wrote:
>
> Hi, I am trying to figure out what recommender or a combination of other
> techniques to use.
> I have a system where a user is presented with a list of pre-defined (max
> 50 products) when he/she is viewing a video clip from a finite (not huge)
> set.    The relationship between video clip and what products to show is
> statically defined and can change from time to time.   We collect clicks
> and resulting buys for these products.   What we are trying to do is to
> come up with the recommender engine that would display top 4-5 products
> that a user would most likely click on and possibly buy (limited to the
> pre-defined list of up to 50 or so products for that particular video clip).
> We can think of a set of these pre-defined products as a category, but a
> product can appear in multiple categories.
>
> What would be the best approach?
>
> Thanks.
>
> ________________________________
> The information contained in this electronic transmission is intended only
> for the use of the recipient and may be confidential and privileged.
> Unauthorized use, disclosure, or reproduction is strictly prohibited and
> may be unlawful. If you have received this electronic transmission in
> error, please notify the sender immediately.
>