Reading the Mind of the Internet: A Study of Collaborative Recommendation Algorithms

Recommender systems are emerging as a key way to manage data on the Internet. In this paper, an overview of different recommender systems is presented, including collaborative, content-based, knowledge-based, and hybrid algorithms. Each of these methods is examined for strengths, weaknesses, and preferred content. Based on this research, the design and implementation of a real-world recommender is explained as a proof of concept. The Webcomic Companion, an online webcomic recommendation system, is outlined, including design, implementation, and testing results. The main components of the system are a PHP website, a MySQL database, and a recommendation algorithm. The key algorithm in the website is an item-based collaborative system augmented with content-based features. On a small test dataset, the algorithm was found to have an average 33% success rate. Also discussed are the ethical consequences of recommendation systems, including privacy and data diversity