RS TV Show Recommendation

This project was developed in 2010, since no competing services were available at the time. The service quickly became a resource hog due to its popularity, and so unfortunately, I had to shut it down. Meanwhile, recommendation systems got more attention, a few competing services now exist with better algorithms, and as such this project may not make a comeback. It is now deprecated. This page only exists for archival purposes.

Project summary

This is a TV Show recommendation engine. You enter the name of the TV shows you're watching or have watched, and the system will search in a huge database for users with similar tastes, then it will recommend the TV shows they watch. Enter as many names as you can remember, because the more names you enter the higher the accuracy is. As a bonus, the engine will calculate how much you spent watching TV shows.

How the idea began

This idea has started out while I was playing with pushing MySQL to its very limits. I created a simple database structure, I filled in some dummy data, then I created a complex query, which in the end could recommend TV shows using 3 tables, lots of JOINs and also subqueries.

Since the hard work was already done, it would've been a waste not to release it to the public. However, to work, it needed actual data, so I made a post on reddit – and several other forums – asking users to input their data. Under 24 hours I collected 1,000 lists. It was good enough, but not all shows were covered, so I tried writing an email to showRSS and they agreed to provide their user data, so I got about 25,000 more lists.

Temporary erroneous results

For about two months, after the new blacklist feature was introduced, one name typo in the SQL query caused the query to exclude every single TV show that was blacklisted by one of the users.

In theory, this would leave only TV shows which every user loves – or at least no user blacklisted. In practice, these were the rare and old shows from the database that have not yet been blacklisted.