MRR

#evaluation #ranking #retrieval #recommendation

MRR (Mean Reciprocal Rank) helps evaluate the quality of recommendation and information retrieval systems. Simply put, it helps understand the average position of the first relevant item across all user lists.

To calculate MRR, you take the reciprocal of the rank of the first relevant item and average this value across all queries or users.

$\begin{array}{rl}\text{with }U& =\text{the total number of users or queries in the evaluated dataset}\\ ran{k}_i& =\text{the rank position of the }first\text{ relevant document for the }{i}^{t}h\text{ user or query.}\end{array}$

More details here:

• https://www.evidentlyai.com/ranking-metrics/evaluating-recommender-systems#mrr
• https://www.evidentlyai.com/ranking-metrics/mean-reciprocal-rank-mrr
• https://en.wikipedia.org/wiki/Mean_reciprocal_rank