feature_selector: Feature Selection Wrapper

In creditmodel: Toolkit for Credit Modeling, Analysis and Visualization

Description

feature_selector This function uses four different methods (IV, PSI, correlation, xgboost) in order to select important features.The correlation algorithm must be used with IV.

Usage

feature_selector(
  dat_train,
  dat_test = NULL,
  x_list = NULL,
  target = NULL,
  pos_flag = NULL,
  occur_time = NULL,
  ex_cols = NULL,
  filter = c("IV", "PSI", "XGB", "COR"),
  cv_folds = 1,
  iv_cp = 0.01,
  psi_cp = 0.5,
  xgb_cp = 0,
  cor_cp = 0.98,
  breaks_list = NULL,
  hopper = FALSE,
  vars_name = TRUE,
  parallel = FALSE,
  note = TRUE,
  seed = 46,
  save_data = FALSE,
  file_name = NULL,
  dir_path = tempdir(),
  ...
)

Arguments

dat_train	A data.frame with independent variables and target variable.
dat_test	A data.frame of test data. Default is NULL.
x_list	Names of independent variables.
target	The name of target variable.
pos_flag	The value of positive class of target variable, default: "1".
occur_time	The name of the variable that represents the time at which each observation takes place.
ex_cols	A list of excluded variables. Regular expressions can also be used to match variable names. Default is NULL.
filter	The methods for selecting important and stable variables.
cv_folds	Number of cross-validations. Default: 5.
iv_cp	The minimum threshold of IV. 0 < iv_i ; 0.01 to 0.1 usually work. Default: 0.02
psi_cp	The maximum threshold of PSI. 0 <= psi_i <=1; 0.05 to 0.2 usually work. Default: 0.1
xgb_cp	Threshold of XGB feature's Gain. 0 <= xgb_cp <=1. Default is 1/number of independent variables.
cor_cp	Threshold of correlation between features. 0 <= cor_cp <=1; 0.7 to 0.98 usually work. Default is 0.98.
breaks_list	A table containing a list of splitting points for each independent variable. Default is NULL.
hopper	Logical.Filtering screening. Default is FALSE.
vars_name	Logical, output a list of filtered variables or table with detailed IV and PSI value of each variable. Default is FALSE.
parallel	Logical, parallel computing. Default is FALSE.
note	Logical.Outputs info. Default is TRUE.
seed	Random number seed. Default is 46.
save_data	Logical, save results in locally specified folder. Default is FALSE.
file_name	The name for periodically saved results files. Default is "select_vars".
dir_path	The path for periodically saved results files. Default is "./variable"
...	Other parameters.

Value

A list of selected features

See Also

psi_iv_filter, xgb_filter, gbm_filter

Examples

feature_selector(dat_train = UCICreditCard[1:1000,c(2,8:12,26)],
                      dat_test = NULL, target = "default.payment.next.month",
                      occur_time = "apply_date", filter = c("IV", "PSI"),
                      cv_folds = 1, iv_cp = 0.01, psi_cp = 0.1, xgb_cp = 0, cor_cp = 0.98,
                      vars_name = FALSE,note = FALSE)

Example output

Package 'creditmodel' version 1.2.7
  Feature    IV   PSI
1   PAY_0 1.019 0.032
2   PAY_2 0.467 0.022
3   PAY_3 0.419 0.017
4   PAY_4 0.255 0.014
5   PAY_5 0.325 0.018
Warning message:
In train_test_split(dat = dat_train, split_type = "OOT", prop = 0.7,  :
  apply_date is  not date or time, unable to use OOT , split random.

creditmodel documentation built on Jan. 7, 2022, 5:06 p.m.