A Study of Evaluation Metrics for Recommender Algorithms

There are inherent problems with evaluating the accuracy of recommender systems. Commonly-used metrics for recommender systems depend on the number of recommendations produced and the number of hidden items withheld, making it difficult to directly compare one system with another. In this paper we compare recommender algorithms using two datasets; the standard MovieLens set and an e-commerce dataset that has implicit ratings based on browsing behaviour. We introduce a measure that aids in the comparison and show how to compare results with baseline predictions based on random recommendation selections