average_precision()
is an alternative to pr_auc()
that avoids any
ambiguity about what the value of precision
should be when recall == 0
and there are not yet any false positive values (some say it should be 0
,
others say 1
, others say undefined).
It computes a weighted average of the precision values returned from
pr_curve()
, where the weights are the increase in recall from the previous
threshold. See pr_curve()
for the full curve.
average_precision(data, ...) # S3 method for data.frame average_precision(data, truth, ..., estimator = NULL, na_rm = TRUE) average_precision_vec(truth, estimate, estimator = NULL, na_rm = TRUE, ...)
data  A 

...  A set of unquoted column names or one or more

truth  The column identifier for the true class results
(that is a 
estimator  One of 
na_rm  A 
estimate  If 
A tibble
with columns .metric
, .estimator
,
and .estimate
and 1 row of values.
For grouped data frames, the number of rows returned will be the same as the number of groups.
For average_precision_vec()
, a single numeric
value (or NA
).
The computation for average precision is a weighted average of the precision
values. Assuming you have n
rows returned from pr_curve()
, it is a sum
from 2
to n
, multiplying the precision value p_i
by the increase in
recall over the previous threshold, r_i  r_(i1)
.
$$AP = \sum (r_{i}  r_{i1}) * p_i$$
By summing from 2
to n
, the precision value p_1
is never used. While
pr_curve()
returns a value for p_1
, it is technically undefined as
tp / (tp + fp)
with tp = 0
and fp = 0
. A common convention is to use
1
for p_1
, but this metric has the nice property of avoiding the
ambiguity. On the other hand, r_1
is well defined as long as there are
some events (p
), and it is tp / p
with tp = 0
, so r_1 = 0
.
When p_1
is defined as 1
, the average_precision()
and roc_auc()
values are often very close to one another.
Macro and macroweighted averaging is available for this metric.
The default is to select macro averaging if a truth
factor with more
than 2 levels is provided. Otherwise, a standard binary calculation is done.
See vignette("multiclass", "yardstick")
for more information.
There is no common convention on which factor level should
automatically be considered the "event" or "positive" result.
In yardstick
, the default is to use the first level. To
change this, a global option called yardstick.event_first
is
set to TRUE
when the package is loaded. This can be changed
to FALSE
if the last level of the factor is considered the
level of interest by running: options(yardstick.event_first = FALSE)
.
For multiclass extensions involving onevsall
comparisons (such as macro averaging), this option is ignored and
the "one" level is always the relevant result.
pr_curve()
for computing the full precision recall curve.
pr_auc()
for computing the area under the precision recall curve using
the trapezoidal rule.
Other class probability metrics:
gain_capture()
,
mn_log_loss()
,
pr_auc()
,
roc_auc()
,
roc_aunp()
,
roc_aunu()
#  # Two class example # `truth` is a 2 level factor. The first level is `"Class1"`, which is the # "event of interest" by default in yardstick. See the Relevant Level # section above. data(two_class_example) # Binary metrics using class probabilities take a factor `truth` column, # and a single class probability column containing the probabilities of # the event of interest. Here, since `"Class1"` is the first level of # `"truth"`, it is the event of interest and we pass in probabilities for it. average_precision(two_class_example, truth, Class1)#> # A tibble: 1 x 3 #> .metric .estimator .estimate #> <chr> <chr> <dbl> #> 1 average_precision binary 0.947#  # Multiclass example # `obs` is a 4 level factor. The first level is `"VF"`, which is the # "event of interest" by default in yardstick. See the Relevant Level # section above. data(hpc_cv) # You can use the col1:colN tidyselect syntax library(dplyr) hpc_cv %>% filter(Resample == "Fold01") %>% average_precision(obs, VF:L)#> # A tibble: 1 x 3 #> .metric .estimator .estimate #> <chr> <chr> <dbl> #> 1 average_precision macro 0.617# Change the first level of `obs` from `"VF"` to `"M"` to alter the # event of interest. The class probability columns should be supplied # in the same order as the levels. hpc_cv %>% filter(Resample == "Fold01") %>% mutate(obs = relevel(obs, "M")) %>% average_precision(obs, M, VF:L)#> # A tibble: 1 x 3 #> .metric .estimator .estimate #> <chr> <chr> <dbl> #> 1 average_precision macro 0.617#> # A tibble: 10 x 4 #> Resample .metric .estimator .estimate #> <chr> <chr> <chr> <dbl> #> 1 Fold01 average_precision macro 0.617 #> 2 Fold02 average_precision macro 0.625 #> 3 Fold03 average_precision macro 0.699 #> 4 Fold04 average_precision macro 0.685 #> 5 Fold05 average_precision macro 0.625 #> 6 Fold06 average_precision macro 0.656 #> 7 Fold07 average_precision macro 0.617 #> 8 Fold08 average_precision macro 0.659 #> 9 Fold09 average_precision macro 0.632 #> 10 Fold10 average_precision macro 0.611# Weighted macro averaging hpc_cv %>% group_by(Resample) %>% average_precision(obs, VF:L, estimator = "macro_weighted")#> # A tibble: 10 x 4 #> Resample .metric .estimator .estimate #> <chr> <chr> <chr> <dbl> #> 1 Fold01 average_precision macro_weighted 0.750 #> 2 Fold02 average_precision macro_weighted 0.745 #> 3 Fold03 average_precision macro_weighted 0.794 #> 4 Fold04 average_precision macro_weighted 0.757 #> 5 Fold05 average_precision macro_weighted 0.740 #> 6 Fold06 average_precision macro_weighted 0.747 #> 7 Fold07 average_precision macro_weighted 0.751 #> 8 Fold08 average_precision macro_weighted 0.759 #> 9 Fold09 average_precision macro_weighted 0.714 #> 10 Fold10 average_precision macro_weighted 0.742# Vector version # Supply a matrix of class probabilities fold1 < hpc_cv %>% filter(Resample == "Fold01") average_precision_vec( truth = fold1$obs, matrix( c(fold1$VF, fold1$F, fold1$M, fold1$L), ncol = 4 ) )#> [1] 0.6173363