The mean of the precisions measured at each rank holding a relevant
item, divided by \(\min(|rel|, k)\). The competing convention
divides by \(|rel|\); it is returned as ap_over_nrel so the two
are never silently swapped.
Usage
MapMet(pred_rank, relevant, k)
Arguments
- pred_rank
A ranking, or a list of rankings.
- relevant
Relevant ids, or a list of such vectors.
- k
Cutoff rank.
Value
Named list with estimate, ap, ap_over_nrel,
n_queries, k, method.
References
Standard IR metric; no single owning source.
Examples
MapMet(c(1, 2, 3, 4, 5), c(2, 4, 9), 5)
#> $estimate
#> [1] 0.3333333
#>
#> $ap
#> [1] 0.3333333
#>
#> $ap_over_nrel
#> [1] 0.3333333
#>
#> $n_queries
#> [1] 1
#>
#> $k
#> [1] 5
#>
#> $method
#> [1] "mean average precision at k"
#>
