A recommender system, or a recommendation system, is a subclass of information filtering systems that seeks to predict the "rating" or "preference" a user would give to an item. Here we have built a program to do the same and predict user ratings (scaling from 1 to 5) on different movies using the user's previous ratings.

You can get the dataset from this website. Here the "MovieLens 1M Dataset" has been used to test the algorithms. Just download and unzip it in the Data folder.

• User-Based: In this method, the program tries finding users with similar movie tastes (users who have rated movies similarly to our user) using KNN. Consequently, predicts the new movie depending on how the other similar users have predicted the same movie.

• Item-Based: In this method, the program tries finding movies with similar user scores (movies which have been rated by users similar to our movie) using KNN. Consequently, predicts the new rating depending on how the other similar movies have been rated by other users.

• Content-Based: Here we use KNN to predict the score the user will give to a movie depending on how he has rated other movies with similar genres (a user might have a big liking to a specific genre which is considered the content of the movies).

Check main.py to see demo.

• Araz.G Shilabin (code)
• Ali Najafi (paper)