Description
Evaluation Metrics for Implicit-Feedback Recommender Systems.
Description
Calculates evaluation metrics for implicit-feedback recommender systems that are based on low-rank matrix factorization models, given the fitted model matrices and data, thus allowing to compare models from a variety of libraries. Metrics include P@K (precision-at-k, for top-K recommendations), R@K (recall at k), AP@K (average precision at k), NDCG@K (normalized discounted cumulative gain at k), Hit@K (from which the 'Hit Rate' is calculated), RR@K (reciprocal rank at k, from which the 'MRR' or 'mean reciprocal rank' is calculated), ROC-AUC (area under the receiver-operating characteristic curve), and PR-AUC (area under the precision-recall curve). These are calculated on a per-user basis according to the ranking of items induced by the model, using efficient multi-threaded routines. Also provides functions for creating train-test splits for model fitting and evaluation.