Recommendation Systems and Netflix Prize

To improve customer experience, Netflix, an online DVD rental company, tries to recommend relevant movies to its customers. Netflix and Amazon.com's recommendation systems, according to some sources, are among the most sophisticated in the world. On Oct 2, 2006, Netflix announced the Netflix Prize and challenged the world, with a $1 million award, to devise an algorithm that can acheive a 10% improvement in accuracy over its current algorithm within 5 years. The data set for the competition contains about 1GB of data (which is only a subset of the whole data set) and Netflix' algorithm usually takes days to train its recommendation system. The entry from the most accurate contestant so far has made about 6.5% improvement. Hence, the problem is large as well as difficult. We are going to study some of the research papers in recommendation systems and devise algorithms to enter the competition.

Term paper

• CSE 4403: 5-7 pages
• CSE 5801: 10-12 pages

Prerequisites for CSE 4403 (undergraduate students)

• 3.5 or above in GPA and
• B or above in Data Structures (CSE 2010) and
• B or above in one of the following:
  • Operating Systems (CSE 4001) or
  • Analysis of Algorithms (CSE 4081) or
  • Artificial Intelligence (CSE 4301)
• or instructor consent in exceptional cases

Prerequisites for CSE 5801 (graduate students)

• 3.75 or above in GPA and
• B or above in one of the following:
  • Analysis of Algorithms (CSE 5211) or
  • Distributed Computing (CSE 5241) or
  • Artificial Intelligence (CSE 5290) or
  • Machine Learning (CSE 5693)
• or instructor consent in exceptional cases

Resources

• And if You Liked the Movie, a Netflix Contest May Reward You Handsomely (New York Times; Oct 2, 2006)
• The $1 Million Netflix Challenge [including an interview with a Netflix VP] (Technology Review; Oct 6, 2006)