Precomputation

[Young <wo...@126.com>]

Hi all,
I am currently working on building a recommendation engine for one business website. I am holding 3 tables, --"favorite items table", "rated items table" and "purchased items tables".
I want to give recommendations based on browsing history. I am new to this field and I do not know which should I pre-compute offline and which should be computed online. 
Could someone give me some suggestions or any papers?
 
Thanks very much.
--Young

Re: Re: Precomputation

[Sean Owen <sr...@gmail.com>]

That's reasonably large. How many rows are in the ratings table? If
it's more than about 100M I think you will need distributed
approaches, which are necessarily offline. So that would answer that
question.

2010/7/3 Young <wo...@126.com>:
> Thanks for your reply. There are about 1 million items in these website and now the users are expected to be 100,000. Is there any problem if I do not do any pre-computation?
>

Re:Re: Precomputation

[Young <wo...@126.com>]

Thanks for your reply. There are about 1 million items in these website and now the users are expected to be 100,000. Is there any problem if I do not do any pre-computation?





>The simplest thing to do is compute on-line, in real time, unless you
>have a reason to do something harder and more complex. And that reason
>would probably be scale. How big are we talking?
>
>There are some links here:
>https://cwiki.apache.org/confluence/display/MAHOUT/Recommender+Documentation
>
>though they may not pertain to the question of doing it online or offline.
>
>2010/7/3 Young <wo...@126.com>:
>> Hi all,
>> I am currently working on building a recommendation engine for one business website. I am holding 3 tables, --"favorite items table", "rated items table" and "purchased items tables".
>> I want to give recommendations based on browsing history. I am new to this field and I do not know which should I pre-compute offline and which should be computed online.
>> Could someone give me some suggestions or any papers?
>>
>> Thanks very much.
>> --Young

Re: Precomputation

[Sean Owen <sr...@gmail.com>]

The simplest thing to do is compute on-line, in real time, unless you
have a reason to do something harder and more complex. And that reason
would probably be scale. How big are we talking?

There are some links here:
https://cwiki.apache.org/confluence/display/MAHOUT/Recommender+Documentation

though they may not pertain to the question of doing it online or offline.

2010/7/3 Young <wo...@126.com>:
> Hi all,
> I am currently working on building a recommendation engine for one business website. I am holding 3 tables, --"favorite items table", "rated items table" and "purchased items tables".
> I want to give recommendations based on browsing history. I am new to this field and I do not know which should I pre-compute offline and which should be computed online.
> Could someone give me some suggestions or any papers?
>
> Thanks very much.
> --Young