A Movie Maker's Guide: Movie Recommendation and Optimization

Abstract

The goal of this paper is to optimize the movie's quality through production process and recommendation system. There are three main factors to consider which includes: Team, Theme, and Time. The paper will also look into hybrid features such as whether an actor from the theme is suitable for a certain theme of a movie.

We present extract dataset from several sources such as IMDb and Movielens, and we formulate our optimization problems using different approaches. We present two methods for recommending movies in recommendation systems which are matrix factorization and graph-based algorithms. Apart from the conventional collaborative filtering in matrix factorization which consists of stochastic gradient descent and alternating least square methods, we propose different approaches using network structure of audience and movies they rated to disclose their preferences and rank popular movies. The methods require walk simulations on tripartite graph, and updating importance score iteratively.

We solve optimization problem in the aspect of movie producers using the tools we have developed in recommendation systems. We are optimizing in the aspect of choosing the best team, picking the right movie theme, and scheduling the optimal release time of movies. Team and theme section can be combined in a single problem with iterative method to maximize the user preferences on movies. Lastly, we pick the most appropriate release dates so that our movies do not have conflicts with movies with similar genres from other studios.