Skip to contents

Cumulative Gain, Lift and Response

Source: R/model_metrics.R
gain_lift.Rd

This function calculates cumulative gain, lift, and response values for a predictive score of a specific target. You can use the mplot_gain() function to create a plot.

Usage

gain_lift(
  tag,
  score,
  target = "auto",
  splits = 10,
  plot = FALSE,
  quiet = FALSE
)

Arguments

tag

Vector. Real known label

score

Vector. Predicted value or model's result

target

Value. Which is your target positive value? If set to 'auto', the target with largest mean(score) will be selected. Change the value to overwrite. Only used when binary categorical model.

splits

Integer. Number of percentiles to split the data

plot

Boolean. Plot results? Uses mplot_gain()

quiet

Boolean. Quiet all messages, warnings, recommendations?

Value

data.frame when plot=FALSE or plot when plot=TRUE.

See also

Other Machine Learning: ROC(), conf_mat(), export_results(), h2o_automl(), h2o_predict_API(), h2o_predict_MOJO(), h2o_predict_binary(), h2o_predict_model(), h2o_selectmodel(), impute(), iter_seeds(), lasso_vars(), model_metrics(), model_preprocess(), msplit()

Other Model metrics: ROC(), conf_mat(), errors(), loglossBinary(), model_metrics()

Examples

data(dfr) # Results for AutoML Predictions
head(dfr$class2)
#>     tag    scores
#> 1  TRUE 0.3155498
#> 2  TRUE 0.8747599
#> 3  TRUE 0.8952823
#> 4 FALSE 0.0436517
#> 5  TRUE 0.2196593
#> 6 FALSE 0.2816101

# Results for Binomial Model
gain_lift(dfr$class2$tag, dfr$class2$scores, target = "FALSE")
#> Target value: FALSE
#> # A tibble: 10 × 10
#>    percentile value random target total  gain optimal  lift response score
#>    <fct>      <chr>  <dbl>  <int> <int> <dbl>   <dbl> <dbl>    <dbl> <dbl>
#>  1 1          FALSE   10.1     27    27  16.4    16.4 62.4     16.4  93.9 
#>  2 2          FALSE   20.1     23    27  30.3    32.7 50.4     13.9  92.5 
#>  3 3          FALSE   30.2     26    27  46.1    49.1 52.4     15.8  90.7 
#>  4 4          FALSE   39.9     23    26  60      64.8 50.3     13.9  86.8 
#>  5 5          FALSE   50       22    27  73.3    81.2 46.7     13.3  80.4 
#>  6 6          FALSE   60.1     18    27  84.2    97.6 40.2     10.9  68.8 
#>  7 7          FALSE   69.8     14    26  92.7   100   32.9      8.48 54.0 
#>  8 8          FALSE   80.2      8    28  97.6   100   21.6      4.85 19.8 
#>  9 9          FALSE   90.3      2    27  98.8   100    9.40     1.21  6.66
#> 10 10         FALSE  100        2    26 100     100    0        1.21  1.76
gain_lift(dfr$class2$tag, dfr$class2$scores, target = "TRUE", splits = 5)
#> Target value: TRUE
#> # A tibble: 5 × 10
#>   percentile value random target total  gain optimal  lift response score
#>   <fct>      <chr>  <dbl>  <int> <int> <dbl>   <dbl> <dbl>    <dbl> <dbl>
#> 1 1          TRUE    20.5     50    55  48.5    53.4 137.     48.5  79.7 
#> 2 2          TRUE    39.9     31    52  78.6   100    97.0    30.1  31.5 
#> 3 3          TRUE    60.1     14    54  92.2   100    53.5    13.6  13.3 
#> 4 4          TRUE    79.9      5    53  97.1   100    21.6     4.85  7.49
#> 5 5          TRUE   100        3    54 100     100     0       2.91  3.19