§ WOE transformation is now available to all the classifier groups. If enabled, classifiers can use transformed WOE features instead of the read-to-use features in the model.
§ WOE specific arguments for any classifier are as follows:
1 woe_transform: This is a Boolean expression and has two values: True or False. By default, this value is False.
2 woe.bin.method: This is a string value that specifies the type of binning to be used for numeric variables before woe transformations, and converts numeric to discrete values. Values are quantile, interval, or auto. For example, woe_bin_method = 'quantile'.
3 woe.n.tile: This value specifies the number of bins to be created when converting numeric to discrete data. For example, woe_n_tile = 10.
|
WARNING |
When fitting XGB, if you foresee future level changes in a categorical variable, you must implement transformations to make sure 1-hot (?) encoding is not used. Doing this breaks the model. To prevent this, enable WOE in the XBG classifier group. |
To add values for a technique, execute the commands as shown in the following paragraph:
|
%python
clsf_grid = [ {"models": ["WOELR", "LR", "ADB", "MLP", "XGB", "GB"]},
{"models": ["XGB"], "params":{'n_estimators': [10,20]}, "ctrls": {"feature_include": None,"require_test": True}} ] |