Detection rate is the proportion of all observations that are true positives.
Usage
detection_rate(data, ...)
# S3 method for class 'data.frame'
detection_rate(
data,
truth,
estimate,
estimator = NULL,
na_rm = TRUE,
case_weights = NULL,
event_level = yardstick_event_level(),
...
)
detection_rate_vec(
truth,
estimate,
estimator = NULL,
na_rm = TRUE,
case_weights = NULL,
event_level = yardstick_event_level(),
...
)Arguments
- data
Either a
data.framecontaining the columns specified by thetruthandestimatearguments, or atable/matrixwhere 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, afactorvector.- estimate
The column identifier for the predicted class results (that is also
factor). As withtruththis can be specified different ways but the primary method is to use an unquoted variable name. For_vec()functions, afactorvector.- 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
logicalvalue indicating whetherNAvalues 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 oftruthto 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 detection_rate_vec(), a single numeric value (or NA).
Details
Suppose a 2x2 table with notation:
| Reference | ||
| Predicted | Positive | Negative |
| Positive | A | B |
| Negative | C | D |
The formula used here is:
$$\text{Detection Rate} = \frac{A}{A + B + C + D}$$
Detection rate is a metric that should be maximized. The output ranges from 0 to 1. For binary data, its maximum is the prevalence of the positive class.
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.
Macro and micro averaging equal accuracy divided by the number of class levels. Macro-weighted averaging weights per-class detection rates by observed class frequencies. Because each per-class rate is divided by the total number of observations, the maximum of the macro and micro estimates is 1 divided by the number of class levels. Case weights apply to both rates and frequencies.
See also
All class metrics, detection_prevalence()
Other class metrics:
accuracy(),
bal_accuracy(),
detection_prevalence(),
f_meas(),
fall_out(),
j_index(),
kap(),
markedness(),
mcc(),
miss_rate(),
npv(),
ppv(),
precision(),
recall(),
roc_dist(),
sedi(),
sens(),
spec()
Examples
# Two class
data("two_class_example")
detection_rate(two_class_example, truth, predicted)
#> # A tibble: 1 × 3
#> .metric .estimator .estimate
#> <chr> <chr> <dbl>
#> 1 detection_rate binary 0.454
# Multiclass
library(dplyr)
data(hpc_cv)
hpc_cv |>
filter(Resample == "Fold01") |>
detection_rate(obs, pred)
#> # A tibble: 1 × 3
#> .metric .estimator .estimate
#> <chr> <chr> <dbl>
#> 1 detection_rate macro 0.182
# Groups are respected
hpc_cv |>
group_by(Resample) |>
detection_rate(obs, pred)
#> # A tibble: 10 × 4
#> Resample .metric .estimator .estimate
#> <chr> <chr> <chr> <dbl>
#> 1 Fold01 detection_rate macro 0.182
#> 2 Fold02 detection_rate macro 0.178
#> 3 Fold03 detection_rate macro 0.189
#> 4 Fold04 detection_rate macro 0.178
#> 5 Fold05 detection_rate macro 0.178
#> 6 Fold06 detection_rate macro 0.174
#> 7 Fold07 detection_rate macro 0.169
#> 8 Fold08 detection_rate macro 0.180
#> 9 Fold09 detection_rate macro 0.168
#> 10 Fold10 detection_rate macro 0.175
# Weighted macro averaging
hpc_cv |>
group_by(Resample) |>
detection_rate(obs, pred, estimator = "macro_weighted")
#> # A tibble: 10 × 4
#> Resample .metric .estimator .estimate
#> <chr> <chr> <chr> <dbl>
#> 1 Fold01 detection_rate macro_weighted 0.311
#> 2 Fold02 detection_rate macro_weighted 0.306
#> 3 Fold03 detection_rate macro_weighted 0.315
#> 4 Fold04 detection_rate macro_weighted 0.302
#> 5 Fold05 detection_rate macro_weighted 0.302
#> 6 Fold06 detection_rate macro_weighted 0.298
#> 7 Fold07 detection_rate macro_weighted 0.289
#> 8 Fold08 detection_rate macro_weighted 0.303
#> 9 Fold09 detection_rate macro_weighted 0.286
#> 10 Fold10 detection_rate macro_weighted 0.297
# Vector version
detection_rate_vec(
two_class_example$truth,
two_class_example$predicted
)
#> [1] 0.454
# Making Class2 the "relevant" level
detection_rate_vec(
two_class_example$truth,
two_class_example$predicted,
event_level = "second"
)
#> [1] 0.384
