Actually it is very easy to mix item and user recs, just put both item and user 
ids in the query.

You can also boost item over user or user over item since they both allow 
special boosts in the query. No need to consult the EventServer.

Note here: http://actionml.com/docs/ur_queries 
<http://actionml.com/docs/ur_queries> That the user-id and item-id can be 
supplied simultaneously and there are user and item boosts. In the UR 0.6 (soon 
to be released) we add item-set queries, to find items similar to a group of 
items, this is used for complimentary purchases or wishlist recommendations 
(though with slightly different training setups)


On Mar 23, 2017, at 8:56 AM, Marius Rabenarivo <[email protected]> 
wrote:

Hi,

To add some informations to Pat's authoritative response :)

You can use Universal Recommener (UR) for this.
With purchase as your primary indicator (http://actionml.com/docs/ur_config 
<http://actionml.com/docs/ur_config>)

By sending item ID in the query you can do item-item similarity 
(http://actionml.com/docs/ur_queries <http://actionml.com/docs/ur_queries>)

If you want to do hybrid method : with both user based collaborative filtering 
and item-item similarity,
I think you can do it by boosting (bias > 1) items with similar properties as 
the item being displayed or for
items in the users profile in the query.

For this last, you will have to query the event server for events relative to a 
particular user. In the documentation
it is said that they should not be supported by SDK nor used by real 
application under normal circumstances and they are subject to changes.
(http://predictionio.incubator.apache.org/datacollection/eventapi/#debugging-recipes
 
<http://predictionio.incubator.apache.org/datacollection/eventapi/#debugging-recipes>)
I'm interested if someone have another way to do hybrid method.

Regards,

Marius

2017-03-23 18:08 GMT+04:00 Vaghawan Ojha <[email protected] 
<mailto:[email protected]>>:
Hi, 

I've been trying to deploy a recommendation system using 
https://github.com/PredictionIO/template-scala-parallel-universal-recommendation
 
<https://github.com/PredictionIO/template-scala-parallel-universal-recommendation>.
 

I've purchase history of user something like this: 
user_id, product_id and purchase_date, so I will be using user_id and 
product_id to determine the recommendation. I'm not sure if I would be able to 
customize the default even parameter. 

Do you have any suggestions like which template would be more suitable for my 
problem. I don't have data like rating or view state, I only have data about 
user and product they purchased. I need something like item based similarity as 
well as user based item similarity. 

Any help would be great

Thank you
Vaghawan


Reply via email to