The evaluation metrics MRR and NDCG consider the position of relevant items in the recommendation list.

The average precision (AP) metric accounts for relevant items not in the recommendation list, therefore, implicitly factoring in recall.

Mean absolute error (MAE) disproportionally penalizes larger discrepancies between predicted and true ratings, in contrast to RMSE which does not.

NDCG relates DCG to ideal DCG, to compensate for the sparsity in users' utility scores.

Below you find either a definition or an example of the different hybridization paradigms according to Burke. Select the correct one.

Graph-based transitivity operates on a bipartite graph, encoding users and items as nodes, and edges as indicators of interactions.

A recommender system adopting graph-based transitivity considers the recommendation task as a graph analysis problem. First, a bipartite graph of users and items as nodes is constructed. Item i is recommended to user u if there exists a path of length l ≤ M between i and u, and M is an even number.

Tie strength or embeddedness is an edge measure in (social) network analysis, which is defined as the overlap (Jaccard index) between two users' neighborhoods.

Local bridges in a graph (e.g., social network) refer to edges the removal of which results in disconnected graphs.

Node measures, such as centrality, in (social) network analysis can be used to compute an importance score for a user. In a memory-based user-CF for rating prediction, adding this score as a weighting term to the user similarity function makes including a user bias term obsolete.