如何用 pandas 数据框中的 WOE 值替换数据框中所有列的 woe 值

Question

I am trying to replace all the columns in a Pandas data frame with respective woe values.

I calculated woe values in a separate function.

I have variable, bin, binedges, WOE in one data frame and in the main data frame.

I have customer_id and the rest of independent varaibles, i have replace the independent varaible values with respective woe values.

Can any one please help?

Answer 1

You can use the xverse package in python for this.

First of all install the xverse package using Anaconda Prompt:

pip install xverse

Note: I'm also showing how to make bins.

Then import MonotonicBinning from the xverse package in your notebook and make bins.

from xverse.transformer import MonotonicBinning

clf = MonotonicBinning()
clf.fit(X, y)
output_bins = clf.bins

Where X is the set of features(of which you want to replace by woe values) as pandas Dataframe and y is the target variable in form of an array

Now store the bins in a separate dataset with the same column names:

X1 = clf.transform(X)

Now import WOE from the xverse package

from xverse.transformer import WOE
clf1 = WOE()
clf1.fit(X1, y)

X2 = clf1.transform(X1)

X2 is the required dataframe of features replaced by their respective woe values

Answer 2

You can use XVERSE.

Step-1: Feature Subset select a subset of features from the dataset. A list of features should be provided to subset.

from xverse.feature_subset import FeatureSubset
numerical_features = list(df._get_numeric_data().columns)
categorical_features = list(df.columns.difference(numerical_features))
print(numerical_features)

clf = FeatureSubset(numerical_features) #select only numeric features
df = clf.fit_transform(df) #returns the dataframe with selected features 

Step-2: Split X and Y

from xverse.feature_subset import SplitXY
clf = SplitXY(['target']) #Split the dataset into X and y
X, y = clf.fit_transform(df) #returns features (X) dataset and target(Y) as a numpy array

Step-3: Weight of Evidence

from xverse.transformer import WOE
clf = WOE()
clf.fit(X, y)

Have a look at the Information value of each of the features clf.iv_df

output_woe_bins = clf.woe_bins #future transformation 
output_mono_bins = clf.mono_custom_binning  #future transformation 

Also, Using the custom binning option in the future to score new data - WOE

clf = WOE(woe_bins=output_woe_bins, mono_custom_binning=output_mono_bins) #output_bins was created earlier
out_X = clf.transform(X)

Take some time to have a complete understanding of the Parameters of WOE

feature_names: 'all' or list (default='all')
    list of features to perform WOE transformation. 
    - 'all' (default): All categorical features in the dataset will be used
    - list of features: ['age', 'income',......]

exclude_features: list (default=None)
    list of features to be excluded from WOE transformation.
    - Example - ['age', 'income', .......]

woe_prefix: string (default=None)
    Variable prefix to be used for the column created by WOE transformer. The default value is set 'None'.  

treat_missing: {'separate', 'mode', 'least_frequent'} (default='separate')
    This parameter setting is used to handle missing values in the dataset.
    'separate' - Missing values are treated as a own group (category)
    'mode' - Missing values are combined with the highest frequent item in the dataset
    'least_frequent' - Missing values are combined with the least frequent item in the dataset

woe_bins: dict of dicts(default=None)
    This feature is added as part of future WOE transformations or scoring. If this value is set, then WOE values provided for each of the features here will be used for transformation. Applicable only in the transform method. 
    Dictionary structure - {'feature_name': float list}
    Example - {'education': {'primary' : 0.1, 'tertiary' : 0.5, 'secondary', 0.7}}

monotonic_binning: bool (default=True)
    This parameter is used to perform monotonic binning on numeric variables. If set to False, numeric variables would be ignored.

mono_feature_names: 'all' or list (default='all')
    list of features to perform monotonic binning operation. 
    - 'all' (default): All features in the dataset will be used
    - list of features: ['age', 'income',......]

mono_max_bins: int (default=20)
    Maximum number of bins that can be created for any given variable. The final number of bins created will be less than or equal to this number.

mono_force_bins: int (default=3)
    It forces the module to create bins for a variable, when it cannot find monotonic relationship using "max_bins" option. The final number of bins created will be equal to the number specified.

mono_cardinality_cutoff: int (default=5)
    Cutoff to determine if a variable is eligible for monotonic binning operation. Any variable which has unique levels less than this number will be treated as character variables. At this point no binning operation will be performed on the variable and it will return the unique levels as bins for these variable.

mono_prefix: string (default=None)
    Variable prefix to be used for the column created by monotonic binning. 

mono_custom_binning: dict (default=None)
    Using this parameter, the user can perform custom binning on variables. This parameter is also used to apply previously computed bins for each feature (Score new data).
    Dictionary structure - {'feature_name': float list}
    Example - {'age': [0., 1., 2., 3.]