These functions calculate the precision()
of a measurement system for
finding relevant documents compared to reference results
(the truth regarding relevance). Highly related functions are recall()
and f_meas()
.
Usage
precision(data, ...)
# S3 method for data.frame
precision(
data,
truth,
estimate,
estimator = NULL,
na_rm = TRUE,
case_weights = NULL,
event_level = yardstick_event_level(),
...
)
precision_vec(
truth,
estimate,
estimator = NULL,
na_rm = TRUE,
case_weights = NULL,
event_level = yardstick_event_level(),
...
)
Arguments
- data
Either a
data.frame
containing the columns specified by thetruth
andestimate
arguments, or atable
/matrix
where the true class results should be in the columns of the table.- ...
Not currently used.
- truth
The column identifier for the true class results (that is a
factor
). This should be an unquoted column name although this argument is passed by expression and supports quasiquotation (you can unquote column names). For_vec()
functions, afactor
vector.- estimate
The column identifier for the predicted class results (that is also
factor
). As withtruth
this can be specified different ways but the primary method is to use an unquoted variable name. For_vec()
functions, afactor
vector.- estimator
One of:
"binary"
,"macro"
,"macro_weighted"
, or"micro"
to specify the type of averaging to be done."binary"
is only relevant for the two class case. The other three are general methods for calculating multiclass metrics. The default will automatically choose"binary"
or"macro"
based onestimate
.- na_rm
A
logical
value indicating whetherNA
values should be stripped before the computation proceeds.- case_weights
The optional column identifier for case weights. This should be an unquoted column name that evaluates to a numeric column in
data
. For_vec()
functions, a numeric vector,hardhat::importance_weights()
, orhardhat::frequency_weights()
.- event_level
A single string. Either
"first"
or"second"
to specify which level oftruth
to consider as the "event". This argument is only applicable whenestimator = "binary"
. The default uses an internal helper that defaults to"first"
.
Value
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 precision_vec()
, a single numeric
value (or NA
).
Details
The precision is the percentage of predicted truly relevant results of the total number of predicted relevant results and characterizes the "purity in retrieval performance" (Buckland and Gey, 1994).
When the denominator of the calculation is 0
, precision is undefined. This
happens when both # true_positive = 0
and # false_positive = 0
are true,
which mean that there were no predicted events. When computing binary
precision, a NA
value will be returned with a warning. When computing
multiclass precision, the individual NA
values will be removed, and the
computation will procede, with a warning.
Relevant Level
There is no common convention on which factor level should
automatically be considered the "event" or "positive" result
when computing binary classification metrics. In yardstick
, the default
is to use the first level. To alter this, change the argument
event_level
to "second"
to consider the last level of the factor the
level of interest. For multiclass extensions involving one-vs-all
comparisons (such as macro averaging), this option is ignored and
the "one" level is always the relevant result.
Multiclass
Macro, micro, and macro-weighted 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.
Implementation
Suppose a 2x2 table with notation:
Reference | ||
Predicted | Relevant | Irrelevant |
Relevant | A | B |
Irrelevant | C | D |
The formulas used here are:
$$recall = A/(A+C)$$ $$precision = A/(A+B)$$ $$F_{meas} = (1+\beta^2) * precision * recall/((\beta^2 * precision)+recall)$$
See the references for discussions of the statistics.
References
Buckland, M., & Gey, F. (1994). The relationship between Recall and Precision. Journal of the American Society for Information Science, 45(1), 12-19.
Powers, D. (2007). Evaluation: From Precision, Recall and F Factor to ROC, Informedness, Markedness and Correlation. Technical Report SIE-07-001, Flinders University
Examples
# Two class
data("two_class_example")
precision(two_class_example, truth, predicted)
#> # A tibble: 1 × 3
#> .metric .estimator .estimate
#> <chr> <chr> <dbl>
#> 1 precision binary 0.819
# Multiclass
library(dplyr)
data(hpc_cv)
hpc_cv %>%
filter(Resample == "Fold01") %>%
precision(obs, pred)
#> # A tibble: 1 × 3
#> .metric .estimator .estimate
#> <chr> <chr> <dbl>
#> 1 precision macro 0.637
# Groups are respected
hpc_cv %>%
group_by(Resample) %>%
precision(obs, pred)
#> # A tibble: 10 × 4
#> Resample .metric .estimator .estimate
#> <chr> <chr> <chr> <dbl>
#> 1 Fold01 precision macro 0.637
#> 2 Fold02 precision macro 0.603
#> 3 Fold03 precision macro 0.706
#> 4 Fold04 precision macro 0.658
#> 5 Fold05 precision macro 0.651
#> 6 Fold06 precision macro 0.626
#> 7 Fold07 precision macro 0.562
#> 8 Fold08 precision macro 0.652
#> 9 Fold09 precision macro 0.605
#> 10 Fold10 precision macro 0.625
# Weighted macro averaging
hpc_cv %>%
group_by(Resample) %>%
precision(obs, pred, estimator = "macro_weighted")
#> # A tibble: 10 × 4
#> Resample .metric .estimator .estimate
#> <chr> <chr> <chr> <dbl>
#> 1 Fold01 precision macro_weighted 0.697
#> 2 Fold02 precision macro_weighted 0.690
#> 3 Fold03 precision macro_weighted 0.752
#> 4 Fold04 precision macro_weighted 0.690
#> 5 Fold05 precision macro_weighted 0.705
#> 6 Fold06 precision macro_weighted 0.682
#> 7 Fold07 precision macro_weighted 0.649
#> 8 Fold08 precision macro_weighted 0.702
#> 9 Fold09 precision macro_weighted 0.661
#> 10 Fold10 precision macro_weighted 0.683
# Vector version
precision_vec(
two_class_example$truth,
two_class_example$predicted
)
#> [1] 0.8194946
# Making Class2 the "relevant" level
precision_vec(
two_class_example$truth,
two_class_example$predicted,
event_level = "second"
)
#> [1] 0.8609865