7. The Machine
Learning
Behind
Recommender
Systems
We use historical item-user data to predict
unobserved item-user data
Typically big datasets
i.e. billions of observations
millions of users
tons of items
Numerous Specifically Designed Algorithms
10. Distinction 1:
Considerations
“Oh no! MyTiVo thinks I’m gay”
Jeffrey Zaslow,TheWall Street Journal, December 2002
What I Like versus What I Say I Like
Solution: Use a bit of both implicit and explicit
11. Distinction 2:
Content-Based
versus
Collaborative
Filtering
Supervised learning
Features are extracted from ‘metadata’
Target variable is rating (explicit) or whether the movie will be
watched (implicit)
Genre Director Main
Actor
Year Rating
The Usual
Suspects
Crime Bryan
Singer
Kevin
Spacey
1995
Titanic Drama James
Cameron
Leonardo
DiCaprio
1997
Die Hard Action John
McTiernan
Bruce
Willis
1988
?
16. MyApproach
Start with Open Source Software
Lenskit (Java)
MyMediaLite (C#)
Mahout (Python)
Learn about Recommender Systems and User Base
Scale Up
Cassandra
Akka
17. State-of-the-
Art
We can do predictions really well
Challenges
Cold Start Problem
Context-Aware Recommendations
Social Recommendations
“Merged accounts”
20. Examples of
Things Data
CannotTellUs
Do I feel my privacy invaded?
Am I happy to have American Pie 2 recommended?
Why do people react to recommendations the way they do?
Presentation?
Bad Recommendations?
Choice Overload?
21. We need to do
A/B testing
andUX
measurement
System A System B
22. What did we
learn from
surveys?
Satisfaction =
Recommendation Set
Attractiveness - Choice
Difficulty
More views != Satisfaction
Diversity influences
Satisfaction
Long Lists = Difficult to
Choose
Short Lists = Easier to
Choose, but not enough
choice
Right Balance = Short
Lists of Diverse Items
24. Thank you for
listening!
Some Pointers
Recommender Algorithms
Yehuda Koren, Google
Introduction to Recommender Systems,Coursera/GroupLens
Infrastructure
NetflixTech Blog
A/BTesting
Ron Kohavi, Microsoft Research
User Experience Evaluation in Recommender Systems
Bart Knijnenburg, Clemson University