• API Reference - DominantFrequencyChange

Dominant Frequency Change

This example provides an overview for using and understanding the Dominant Frequency Change check.

Structure:

• What is a Dominant Frequency Change?

• Generate Data

• Run The Check

• Define a Condition

What is a Dominant Frequency Change?

Dominant Frequency Change is a data integrity check which simply checks whether dominant values have increased significantly between test data and train data. Sharp changes in dominant values can indicate a problem with the data collection or data processing pipeline (for example, a sharp increase in a common null or constant value), and will cause the model to fail to generalize well. This check goal is to catch these issues early in the pipeline.

This check compares the dominant values of each feature in the test data to the dominant values of the same feature in the train data. If the ratio of the test to train dominant values is greater than a threshold, the check fails. This threshold can be configured by specifying the ratio_change_thres parameter of the check.

The Definition of a Dominant Value

The dominant value is defined as a value that is frequent in data at least more than dominance_ratio times from the next most frequent value. The dominance_ratio is a configurable parameter of the check.

from deepchecks.tabular.checks import DominantFrequencyChange
from deepchecks.tabular.datasets.classification import iris

Generate data

train_ds, test_ds = iris.load_data(data_format='Dataset', as_train_test=True)

Introducing Duplicates in the Test Data

# make duplicates in the test data
test_ds.data.loc[test_ds.data.index % 2 == 0, 'petal length (cm)'] = 5.1
test_ds.data.loc[test_ds.data.index / 3 > 8, 'sepal width (cm)'] = 2.7

Run The Check

check = DominantFrequencyChange()
check.run(test_ds, train_ds)

The DominantFrequencyChange check is deprecated and will be removed in the 0.11 version.

Define a Condition

check = DominantFrequencyChange()
check.add_condition_ratio_of_change_less_than(0.1)
res = check.run(test_ds, train_ds)
res.show(show_additional_outputs=False)

The DominantFrequencyChange check is deprecated and will be removed in the 0.11 version.

previous

Date Train Test Leakage Overlap

next

Feature Label Correlation Change