SimpleModelComparison#

class SimpleModelComparison[source]#

Compare given model score to simple model score (according to given model type).

Parameters

strategystr, default: ‘most_frequent’

Strategy to use to generate the predictions of the simple model [‘stratified’, ‘uniform’, ‘most_frequent’, ‘tree’].

stratified: randomly draw a label based on the train set label distribution. (Previously ‘random’)
uniform: in regression samples predictions uniformly at random from the y ranges. in classification draws
predictions uniformly at random from the list of values in y.
most_frequent: in regression is mean value, in classification the most common value. (Previously ‘constant’)
tree: runs a simple decision tree.

simple_model_typestr , default: most_frequent

Deprecated. Please use strategy instead.

alternative_scorersDict[str, Callable], default: None

An optional dictionary of scorer title to scorer functions/names. If none given, using default scorers. For description about scorers see Notes below.

max_gainfloat , default: 50

the maximum value for the gain value, limits from both sides [-max_gain, max_gain]

max_depthint , default: 3

the max depth of the tree (used only if simple model type is tree).

random_stateint , default: 42

the random state (used only if simple model type is tree or random).

Notes

Scorers are a convention of sklearn to evaluate a model. See scorers documentation A scorer is a function which accepts (model, X, y_true) and returns a float result which is the score. For every scorer higher scores are better than lower scores.

You can create a scorer out of existing sklearn metrics:

from sklearn.metrics import roc_auc_score, make_scorer

training_labels = [1, 2, 3]
auc_scorer = make_scorer(roc_auc_score, labels=training_labels, multi_class='ovr')
# Note that the labels parameter is required for multi-class classification in metrics like roc_auc_score or
# log_loss that use the predict_proba function of the model, in case that not all labels are present in the test
# set.

Or you can implement your own:

from sklearn.metrics import make_scorer

def my_mse(y_true, y_pred):
    return (y_true - y_pred) ** 2

# Mark greater_is_better=False, since scorers always suppose to return
# value to maximize.
my_mse_scorer = make_scorer(my_mse, greater_is_better=False)

__init__(strategy: str = 'most_frequent', simple_model_type: Optional[str] = None, alternative_scorers: Optional[Dict[str, Callable]] = None, max_gain: float = 50, max_depth: int = 3, random_state: int = 42, **kwargs)[source]#

__new__(*args, **kwargs)#

Methods

`SimpleModelComparison.add_condition`(name, ...)	Add new condition function to the check.
`SimpleModelComparison.add_condition_gain_greater_than`([...])	Add condition - require minimum allowed gain between the model and the simple model.
`SimpleModelComparison.clean_conditions`()	Remove all conditions from this check instance.
`SimpleModelComparison.conditions_decision`(result)	Run conditions on given result.
`SimpleModelComparison.config`()	Return check configuration (conditions' configuration not yet supported).
`SimpleModelComparison.from_config`(conf)	Return check object from a CheckConfig object.
`SimpleModelComparison.metadata`([with_doc_link])	Return check metadata.
`SimpleModelComparison.name`()	Name of class in split camel case.
`SimpleModelComparison.params`([show_defaults])	Return parameters to show when printing the check.
`SimpleModelComparison.remove_condition`(index)	Remove given condition by index.
`SimpleModelComparison.run`(train_dataset, ...)	Run check.
`SimpleModelComparison.run_logic`(context)	Run check.

Examples#

SegmentPerformance.run_logic

SimpleModelComparison.add_condition