Filter query - algorithm

[SoapSurgeon]

Hi,

I have a series of records in MongoDB (although it could be anything for the purpose of this question). Lets say I have a number of products:

Product:
id
name
price
date_added
interestingness

I want to list the 25 most interesting products ordered by date (latest added at the top). I always want to have some products listed so if there aren't any good ones it should choose worse ones. How does this general kind of algorithm work, ideally at a database level.

 

[SoapSurgeon]

Hi,

Thanks for your reply. There will be new 'products' added all the time and I want the most interesting recent ones to show. Being stale is a definite no-no - I don't want a really interesting product that was added a year ago to show on the front page (strange as it may seem!).

I think this is very similar to the facebook feed?

I did think about weighting the products based on date_added. today == 1.0 and +7 days == 0. This would remove any products over a week old. Just wondered if it was a standard problem with a standard solution before I start hacking away

 

[SoapSurgeon]

Hi,

Thanks I've read that link - very interesting but not what I'm after, it might come in handy later though.

What I require is a timeline of products. Let's imagine that 5000 products have been added today. I want to display a subset of those products hand picked for a particular user (the interestingness). The most recent product should be at the top and the user should be able to 'load more' which will select another set of interesting products. An old product should not appear on the list until the user has clicked load more sufficient enough times.

If no 'interesting' products have been added out of the 5000 i still want to show something, so inferior products (but still added today) will be shown.

I cant limit the products to added today, because in the morning you would have very few products - in which case I would want the end of yesterdays to be included...

 

[SoapSurgeon]

It is a value that is calculated on-the-fly based on a number of factors.

For example, if a user has shown interest in a similar product, the product has been purchased/viewed a lot etc will all increase a products interestingness...

This will be calculated during the query. I was thinking a map reduce job might be the way forward, but it isn't quite right.