recommender systems
DESCRIPTION
Sumbul Jahan. Recommender Systems. This world is an over-crowded place. They all want to get our attention. But we need a few of them!. Something which is popular. Something which is of our interest. Something which is liked by people of our community. What we are looking for?. - PowerPoint PPT PresentationTRANSCRIPT
RECOMMENDER SYSTEMSSumbul Jahan
THIS WORLD IS AN OVER-CROWDED PLACE
THEY ALL WANT TO GET OUR ATTENTION
BUT WE NEED A FEW OF THEM!
WHAT WE ARE LOOKING FOR? Something which is popular. Something which is of our interest. Something which is liked by people of
our community.
WHO CAN HELP???
RECOMMENDER SYSTEMS
WHAT IS RECOMMENDER SYSTEMAn information filtering technology, commonly used on e-commerce Web sites that uses a collaborative filtering to present information on items and products that are likely to be of interest to the reader.
http://citationmachine.net/index2.php?reqstyleid=2
WHAT CAN BE RECOMMENDED Advertising
messages Investment choice Restaurants Cafes Music tracks Movies TV programs Books Stores
Tags News articles Future friends Research papers Citations Courses Articles Supermarket goods Products/Services
http://www.slideshare.net/T212/recommender-systems-1311490
TYPES OF RECOMMENDER SYSTEM Content Based
Collaborative Filtering
Knowledge Based
http://www.umiacs.umd.edu/~jimmylin/INFM700-2008 Spring/presentations/recommender_systems.ppt.
CONTENT BASED Recommend items based on user’s
past preferences. Items/content usually denoted by
keywords. Matching “user preferences” with “item
characteristics” … works for textual information.
User profile is the key.
CONTENT BASED - LIMITATIONS Not all content is well represented by
keywords, e.g. images. No profile is available for new users. Unrated items are not shown. Users with thousands of purchases is a
problem.
COLLABORATIVE FILTERING Recommend items based on ratings of
users sharing similar interests. Collaborative filtering systems can
produce personal recommendations by computing the similarity between your preference and the one of other people.
More users, more ratings: better results.
http://pehttp://pespmc1.vub.ac.be/collfilt.html
http://www.bridgewell.com/images_en/ec_03.jpg
COLLABORATIVE FILTERING - LIMITATIONS Different users might use different
scales. Possible solution: weighted ratings, i.e. deviations from average rating.
Finding similar users/user group is not very easy.
No preference is available of new users. No rating is available of new items.
KNOWLEDGE BASED Knowledge of user is linked to
knowledge of products. Conversational interaction used to
establish current user preferences i.e. “more like this”, “less like that”, “none of those” …
No user profiles maintained, preferences drawn through manual interaction
USAGE Netflix
2/3 rented movies are from recommendation.
http://www.shttp://pespmc1.vub.ac.be/collfilt.htmllideshare.net/T212/recommender-systems-1311490
USAGE Google News
More than 38% click-through are due to recommendation.
http://www.slideshare.net/T212/recommender-systems-1311490
USAGE Amazon
35% sales are from recommendation.
http://www.slideshare.net/T212/recommender-systems-1311490