Skip to contents

For each explanatory variable, this function analyzes its relationship with a target variable by calculating statistics, WOE, IV, and other metrics.

Arguments

data

data.frame or data.table - Data to analyze

target

character - Name of the target variable to explain

description_data

character - Description of the dataset (optional)

target_type

character - Type of target: "autoguess" (default), "binary", "categorical", or "numeric"

target_reference_level

any - Reference level for binary/categorical targets (if NULL, will be inferred)

description_target

character - Description of the target variable (optional)

analysis_name

character - Name for the analysis (optional)

select_vars

character vector - Variables to include (if NULL, all columns are considered)

exclude_vars

character vector - Variables to exclude from analysis

nbins

integer - Number of bins for numeric variables (default: 12)

binning_method

character - Method for binning: "quantile" (default), "clustering", or "smart"

naming_conventions

logical - Whether to enforce naming conventions (default: FALSE)

useNA

character - How to handle NAs: "ifany" (default) or "no"

verbose

logical - Whether to print detailed progress information (default: FALSE)

dec

integer - Number of decimals for numeric display (default: 2)

order_label

character - Method for ordering labels in output (default: "auto")

cont_target_trim

numeric - Trimming factor for continuous targets, as percentage between 0 and 1 (default: 0.01)

bxp_factor

numeric - Factor for boxplot whiskers calculation (default: 1.5)

num_as_categorical_nval

integer - Threshold for treating numeric as categorical (default: 5)

autoguess_nrows

integer - Rows to use for variable type detection (default: 1000, 0 means all rows)

woe_alternate_version

character - When to use alternate WOE definition: "if_continuous" (default) or "always"

woe_shift

numeric - Shift value for WOE calculation to prevent issues with 0% or 100% classes (default: 0.01)

woe_post_cluster

logical - Whether to cluster WOE values (default: FALSE)

woe_post_cluster_n

integer - Number of clusters for WOE clustering (default: 6)

smart_quantile_by

numeric - Quantile step for smart binning (default: 0.01)

by_nvars

integer - Number of variables to process in each batch (default: 200)

debug

logical - Whether to print debug information (default: FALSE)

...

Additional parameters passed to targeter_internal

Value

An object of class "targeter" with detailed profiling information

Examples

targeter(adult,target ="ABOVE50K")
#> 
#> INFO:target ABOVE50K detected as type: binary
#> INFO:binary target contains number, automatic chosen level: 1; override using `target_reference_level`
#> 
#> Target profiling object with following properties:
#> 	Target: ABOVE50K  of type: binary  (target level:1 )
#> 	Run on data: adult  the: 2025-03-29
#> 12 profiles available (AGE, FNLWGT, EDUCATIONNUM, HOURSPERWEEK, WORKCLASS...)
#> You can access each crossing using slot $profiles[[__variable__]]. Then on it use `plot` or `summary`
#> You can also directly invoke a global `summary` function on this object.