Content-based Recommender System
Content-base recommender system is the system using content features to predict a user’s rating or preference of a product or a service. Take an online bookstore as an example. You have a movie list and a user list. Each user gives each movie a rating with a value ranged from 0 to 5. There is a feature set for movies. Suppose the feature set includes two features: “romance” and “action”. A give movie has a score for each feature, e.g., 0.9 for romance and 0.1 for action. The goal of content based recommender system is to learn a theta vector for each user so that the theta vector can be leveraged to predict rating for any movie for a particular user. This is basically a liner regression problem. The optimisation objective is to learn the theta matrix to minimise the cost function J(Θ).
