Collaborative Filtering

Hey everyone, I learnt about a really interesting concept: Collaborative filtering.

The above simplified image (which I definitely didn’t spend 15 minutes making) shows what collaborative filtering looks like. The image shows that person A has watched the movies Shrek, Avatar and Minions. Person B has watched Shrek and Avatar, but not minions. Since it seems as if person B has similar interests as person A, it is probable that he might also like minions, and so the algorithm would recommend the movie to person B.

This is only a simplified version, however; The algorithms, in real life, work on a large scale basis. This means that they would only recommend movies (or products, videos, ads, etc.) to people if they are seen to have similar interests to many others, and not just one - which makes the recommendation system much more accurate.

Applications of Collaborative Filtering

Some companies that use collaborative filtering are ones you presumably would have already heard of: Amazon, YouTube, Netflix, and so on… This is how Amazon often recommends you products that you actually like; Netflix shows you what you likely would enjoy (and also how they calculate the percentage chance of you liking a movie/show); YouTube shows you the videos that you are interested in.

All these corporations do an autonomous search for people who are very similar to you, the people who’s shopping interests looks just like yours, the people who watch the same movies and shows as you, and those that watch the same videos as you. They would then recommend you the things that your ‘doppelgangers’ have liked as there’s a good chance you might like those products/movies/videos too.