It was only 2 weeks ago that Socialize held its last hackathon—but we were itching to get in the game again! Our last project, Interests Graphed, was completed in less than 9 hours. What else could we churn out in a days’ work?
Kicking off on Friday at 9:45 am, the team decided on an Amazon-style recommendation engine. Based on similar likes between users, Socialize suggests other things each user might like based on other users’ likes. Phew…say that 3 times fast.
For each given like, we fetched the corresponding user, entity and application, then got all the likes of that user for that application. Once we rounded up all of the other likes for each entity, we sorted by like count to get all the users and their likes—then intersected the likes of each user and application based on entity. The users were then sorted based on the largest overlap, and resorted by entities based on that overlap. (“Yo that’s my name, don’t wear it out” — the word “like”.)
In the end (that is, 4 pm PST) we had a working recommendation engine that further connects users through their app behavior. And saves them hours of sifting through irrelevant content to find something that piques their interest—so 2 weeks ago.