A Group Recommender System for deviantART

deviantART is the world's largest online arts community with a huge number of users and items. They currently recommend art to users only via an algorithmic presentation of "popular" items, and an item-item recommender system presented alongside every individual piece of art. deviantART expects to use recommender systems technology to enhance and leverage the contributions of existing "Groups" of users. The intern will combine time series analysis with collaborative filtering to develop a time-variant Matrix Factorization method and apply this new temporal approach to deviantART’s real data. The intern will develop a prototype of the new proposed method and demonstrate its effectiveness through off-line evaluations on real user group engagement data available from deviantART. Thus this internship will help deviantART with improving the ability of artists and their audiences to connect, and to more generally make using deviantART a more human, social, and ultimately enjoyable experience.

Faculty Supervisor:

Martin Ester

Student:

Xin Wang

Partner:

deviantART Canada Inc.

Discipline:

Computer science

Sector:

Digital media

University:

Simon Fraser University