



Jain-Terkelsen-Hundal Network-Based Inference (JTH-NBI) is a modification of Zhou et al. (2007) network-based algorithm for the projection of a two mode user-object bipartite graph onto a one mode space. The resulting object network can be interpreted as a recommendation network, where edges and edge weights signify artist similarity. From this network, we form recommendation lists for users which suggest objects that a user is most inclined to collect next. JTH-NBI performs better than the Amazon.com’s industry standard item-based collaborative filtering. We test JTH-NBI on the MovieLens data set and data from Hulkshare.com.