Item Based Collaborative Filtering Recommender
In item based collaborative filtering similarity between item’s are used to recommend items to users. In this method if user is looking for item or has liked any item than similar items are recommended to that uers. We use cosine similarity between items to find similarity.
Formula for cosine similarity
In this movie recommendation it uses tf-idf to find similarity between movies. Cosine similarity between user’s selected movie to all other movies are calculted and top 20 results are shown to the users.
To calculate cosine similarity of a movie to all movies for every new movie selection is very inefficient so instead of calculating cosine similarity at runtime we can calculate and cosine similarity of all movies to all movies before hosting and can use that matrix at runtime that will not take too much time.