#05. Cosimo Spera - The evolution of recommendation systems
Details
Program
6:30-7:00 - Networking
7:00-7:30 - The evolution of recommendation systems
7:30-8:00 - Q&A and Discussion
Presentation by Cosimo Spera, BeeBell
Recommendation Systems have been widely studied in the past ten years and many new approaches have been proposed. However, very few people find the suggested recommendations relevant all the time.
I will use two personal experiences to point out at the inconsistency of two of the most popular systems: Amazon and Netflix.
• Example 1: I recently bought a baby stroller on Amazon for a friend of mine who had recently a baby. (My kids are 22 and 16) I even had the purchase to be shipped at her home address. Nevertheless each time I return to the site I get suggested with more strollers and baby diapers. (Obviously the collaborative filtering algorithm is not working properly).
• Example 2: Who does not remember the $1M dollars Netflix challenge on movie recommendations? I love movies and watch them on Netflix all the time, but I found that their recommendations do not reflect the fact that at different time of the day and at different days my mood in not always the same. I do not want to watch a horror movie on Friday night after a week of hard work but these type of movies always come at the top of the list just because my kids watched few times using my account.
In this talk I will try to point out to a new approach that looks at and predict based on the status of three main primary variables (factors): time, location and mood. I will show how this approach can bring a significant breakthrough in this field.
http://photos3.meetupstatic.com/photos/event/6/c/c/1/600_438567841.jpeg