Machine Learning for the Web

Collaborative Filtering methods

This class of methods is based on the idea that any user will like items appreciated by other users similar to them. In simple terms, the fundamental hypothesis is that a user A, who is similar to user B, will likely rate an item as B did rather than in another way. In practice, this concept is implemented by either comparing the taste of different user's and inferring the future rating for a given user using the most similar users taste (memory-based) or by extracting some rating patterns from what the users like (model-based) and trying to predict the future rating following these patterns. All these methods require a large amount of data to work because the recommendations to a given user rely on how many similar ...