Movie Recommender

A backend system that recommends movies to users based on Alternating Least Squares.

It is implemented in Scala by Manos Chatzakis, at EPFL (emmanouil.chatzakis@epfl.ch)

Functionalities

Data Loading: The system parses files of movies and ratings
Analytics: Provides basic analytics about the provided datasets
Aggregator: Provides aggregator functionalities about the movies and ratings. It also supports online data updates for the ratings
Recommender: Baseline and ALS-based recommendations

Movie Recommendations

Given a user and the preferred genres, the system predicts the ratings of the user for movies that match the genres using Alternating Least Squares, and returns the top predictions to the user.

Online Rating Updates

The system supports online rating updates. It calculates the updated rating of a movie using the formula

$$updatedRating = \frac{newRatingsSum + oldRatingsSum}{oldUsers+newUsers}$$

which is an optimized approach to calculate online ratings.

Spark Partitions

The system is designed for Spark. The default number for partitions is 8.

About