Movie Recommendation Based on Graph Traversal Algorithms

Abstract

Media content recommendation is nowadays a common problem. Traditional algorithms based on collaborative filtering require an up-to-date dataset of users and their preferences, which is difficult to gather for huge database of items. Content-based approach suffers from the complex computation of similarity among items. In this paper we propose an approach to recommendation with a focus on the natural change of user's interests in movies. We make use of a graph representation and experimented with modified graph algorithms. We design a representation of the data about movies in a graph structure and a method which uses our data model for recommendation. We propose four recommendation algorithms which are capable to find recommendations based on initial nodes, which selection is based on the user's current interests. We implemented these algorithms and experimentally evaluated them with real users.