Member-only story
An Intro to Collaborative Filtering for Movie Recommendation
Step-by-Step Guide to Recommender System
Recommender system has become a rising topic as we demand more customized contents push to our daily feeds. I guess we are all familiar with the recommended videos on YouTube, and we are all — more than once — the victims of late-night Netflix binge watching.
There are two popular methods in recommender system, collaborative based filtering and content based filtering. Content based filtering makes predictions of what the audience is likely to prefer based on the content properties, e.g. genre, language, video length. Whereas collaborative filtering predicts based on what other similar users also prefer. As the result, collaborative filtering method is leaning towards instance based learning and usually applied by large companies with huge amount of data at hand.
In this article, I will focus on collaborative based filtering and briefly introduce how to make movie recommendation using two algorithms that fall into this category, K Nearest Neighbour (KNN) and Singular Value Decomposition (SVD).
I used the movie dataset from Kaggle to predict recommended movies at individual level.
