Section: New Software and Platforms
Prototype implementation for explaining a set of similar and recommended movies.
In this web-based prototype for similar movies explanation, we propose two types of browsing for: personalized browsing and non personalized browsing. In the non personalized browsing we suppose that we don't have the user profile. Similar movie sublists are ordered only according to their similarity to the selected movie. For the personalized browsing , we select users that have different profiles from our dataset. We give these users names of actors, according to the types of movies they watch. For each user, we compute the predicted ratings using the matrix factorization model. We select pairs of genres to display to each user based on the preferred genres for the user. In our prototype we identify the preferred genres per user based on the most frequent movie genre pairs that the user has already seen. We then organize the recommended movies with a high rating prediction in sublists, according to the user most preferred genre pairs. When a user selects a movie from the sublists of recommended movies, our application suggests the similar movies presented under four sublists with the added list of words. The sublists are personalized for each user by reordering the movies according to the users predicted ratings.