Hi,
I am trying to use Universal Recommender as a template in PredictionIO to
recommend videos for my VOD application. I have my primary event named
"watch" and two secondary events named "view" &"category-pref".
I have listed them as indicators as well in my engine.json file as
seen below:
*"indicators": [*
* {*
* "name":"watch" *
* },*
* {*
* "name": "view",*
* "maxCorrelatorsPerItem": 50*
* },*
* {*
* "name": "category-pref",*
* "maxCorrelatorsPerItem": 50*
* }*
* ]*
My items have some properties which include "genres" and each item's
genres list looks something like this ['Action','Comedy'].
According to my understanding, if user1 watches item1 having genres list as
['genre1','genre2'], than for including this genres list in making
recommendations, what I need to do is send a "watch" event for item1 and
two "category-pref" events for genre1 and genre2 with user1.
So I have two questions:
1. Isn't the algorithm capable of finding correlations by itself for some
user liking specific genres and returning results accordingly?
2. What if I have a large number of secondary events and properties which I
want to record as events to affect my recommendations, do I need to send
all those events for just one primary event occurrence?
Regards,
Sohaib Tanveer
