Recommender systems comparison
This thesis aims to distinguish metrics on recommender systems that can be proved useful to compare them, and performs a comparison between two algorithms of the collaborative filtering family.
Abstract
The last decade Internet has been flooded with information. Information that no one can filter to find what he needs, raw data, videos, music or products. Large retail sites like Amazon developed recommenders systems in order to offer products to their users. The need although is not limited only to the retail area. Web sites like Youtube or Vimeo need to recommend to each user of their, videos that may like to watch next. Facebook is another example of an application utilizing lots of data and offering recommendations on what you may want to read or who may be a friend of yours. Most of the times, a recommender system is not the core functionality of an application. It is through a very useful feature that gives a clear advantage in any business area needed. This thesis aims to distinguish metrics on recommender systems that can be proved useful to compare them. Also, this thesis performs a comparison between two algorithms of the collaborative filtering family. The content-based with focus on items and the machine learning oriented alternating least square (als).