Data drift parameters

How to set custom data drift conditions and thresholds for tabular and text data.

Pre-requisites:

• You know how to generate Reports or Test Suites with default parameters.

• You know how to pass custom parameters for Reports or Test Suites.

• You know how to use Column Mapping to set the input data type.

Default

All Presets, Tests, and Metrics that include data or target (prediction) drift evaluation use the default Data Drift algorithm. It automatically selects an appropriate drift detection method based on the feature type and volume.

You can override the defaults by passing a custom parameter to the chosen Test, Metric, or Preset. You can define the drift detection method, the threshold, or both.

Code example

You can refer to an example How-to-notebook showing how to pass custom drift parameters:

evidently/how_to_specify_stattest_for_a_testsuite.ipynb at main · evidentlyai/evidentlyGitHub

Examples

To set a custom drift method and threshold on the column level:

ColumnDriftMetric(column_name='feature1', stattest='wasserstein', stattest_threshold=0.2) 

If you have a Preset, Test or Metric that checks for drift in multiple columns at the same time, you can set a custom drift method for all columns, all numerical/categorical columns, or for each column individually.

Here is how you set the drift detection method for all categorical columns:

DataDriftPreset(cat_stattest='ks', cat_statest_threshold=0.05)

To set a custom condition for the dataset drift (share of drifting columns in the dataset) in the relevant Metrics or Presets:

DatasetDriftMetric(drift_share=0.7)

Note that this works slightly differently for Tests. To set a custom condition for the dataset drift when you run a relevant Test, you should set a condition for the share of drifted features using standard lt and gt parameters:

TestShareOfDriftedColumns(lt=0.5)

When you set drift threshold for ColumnDriftTest(), you should use stattest_threshold and other parameters the same way as it works in Metrics (not lt and gt).

Tabular drift detection

The following methods and parameters apply to tabular data (as parsed automatically or specified as numerical or categorical columns in the column mapping).

Drift parameters - Tabular

The following drift detection parameters are available in the DataDriftTable(), DatasetDriftMetric(), ColumnDriftMetric(), related Tests, and Presets that contain them.

Parameter	Description
stattest	Defines the drift detection method for a given column (if a single column is tested), or all columns in the dataset (if multiple columns are tested).
stattest_threshold	Sets the drift threshold in a given column or all columns. The threshold meaning varies based on the drift detection method, e.g., it can be the value of a distance metric or a p-value of a statistical test.
drift_share	Defines the share of drifting columns as a condition for Dataset Drift in DatasetDriftMetric or inside a Preset.
cat_stattest cat_stattest_threshold	Sets the drift method and/or threshold for all categorical columns in the dataset.
num_stattest num_stattest_threshold	Sets the drift method and/or threshold for all numerical columns in the dataset.
per_column_stattest per_column_stattest_threshold	Sets the drift method and/or threshold for the listed columns (accepts a dictionary).

How to check available parameters. You can verify which parameters are available for a specific test, metric, or preset in the All tests or All metrics tables or consult the API reference

Drift detection methods - Tabular

To use the following drift detection methods, pass them using the stattest parameter.

StatTest	Applicable to	Drift score
ks Kolmogorov–Smirnov (K-S) test	tabular data only numerical Default method for numerical data, if <= 1000 objects	returns p_value drift detected when p_value < threshold default threshold: 0.05
chisquare Chi-Square test	tabular data only categorical Default method for categorical with > 2 labels, if <= 1000 objects	returns p_value drift detected when p_value < threshold default threshold: 0.05
z Z-test	tabular data only categorical Default method for binary data, if <= 1000 objects	returns p_value drift detected when p_value < threshold default threshold: 0.05
wasserstein Wasserstein distance (normed)	tabular data only numerical Default method for numerical data, if > 1000 objects	returns distance drift detected when distance >= threshold default threshold: 0.1
kl_div Kullback-Leibler divergence	tabular data numerical and categorical	returns divergence drift detected when divergence >= threshold default threshold: 0.1
psi Population Stability Index (PSI)	tabular data numerical and categorical	returns psi_value drift detected when psi_value >= threshold default threshold: 0.1
jensenshannon Jensen-Shannon distance	tabular data numerical and categorical Default method for categorical, if > 1000 objects	returns distance drift detected when distance >= threshold default threshold: 0.1
anderson Anderson-Darling test	tabular data only numerical	returns p_value drift detected when p_value < threshold default threshold: 0.05
fisher_exact Fisher's Exact test	tabular data only categorical	returns p_value drift detected when p_value < threshold default threshold: 0.05
cramer_von_mises Cramer-Von-Mises test	tabular data only numerical	returns p_value drift detected when p_value < threshold default threshold: 0.05
g-test G-test	tabular data only categorical	returns p_value drift detected when p_value < threshold default threshold: 0.05
hellinger Hellinger Distance (normed)	tabular data numerical and categorical	returns distance drift detected when distance >= threshold default threshold: 0.1
mannw Mann-Whitney U-rank test	tabular data only numerical	returns p_value drift detected when p_value < threshold default threshold: 0.05
ed Energy distance	tabular data only numerical	returns distance drift detected when distance >= threshold default threshold: 0.1
es Epps-Singleton tes	tabular data only numerical	returns p_value drift detected when p_value < threshold default threshold: 0.05
t_test T-Test	tabular data only numerical	returns p_value drift detected when p_value < threshold default threshold: 0.05
empirical_mmd Empirical-MMD	tabular data only numerical	returns p_value drift detected when p_value < threshold default threshold: 0.05
TVD Total-Variation-Distance	tabular data only categorical	returns p_value drift detected when p_value < threshold default threshold: 0.05

Text drift detection

Text drift detection applies to columns with raw text data, as specified in column mapping.

Embedding drift detection. If you work with embeddings, you can use Embeddings Drift Detection methods.

Drift parameters - Text

The following text drift detection parameters are available in the DataDriftTable(), DatasetDriftMetric(), ColumnDriftMetric(), related Tests and Presets that contain them.

Parameter	Description
stattest	Defines the drift detection method for a given column that contains text data, or for all columns in the dataset if all columns contain text data.
stattest_threshold	Sets the threshold as a drift detection parameter.
text_stattest	Defines the drift detection method for all text columns in the dataset.
text_stattest_threshold	Sets the threshold as a drift detection parameter.

Drift detection methods - Text

To use the following text drift detection methods, pass them using the stattest parameter.

StatTest	Description	Drift score
perc_text_content_drift Text content drift (domain classifier, with statistical hypothesis testing)	Applies only to text data. Trains a classifier model to distinguish between text in “current” and “reference” datasets. Default for text data when <= 1000 objects.	returns roc_auc of the classifier as a drift_score drift detected when roc_auc > possible ROC AUC of the random classifier at a set percentile threshold sets the percentile of the possible ROC AUC values of the random classifier to compare against default threshold: 0.95 (95th percentile) roc_auc values can be 0 to 1 (typically 0.5 to 1); a higher value means more confident drift detection
abs_text_content_drift Text content drift (domain classifier)	Applies only to text data. Trains a classifier model to distinguish between text in “current” and “reference” datasets. Default for text data when > 1000 objects.	returns roc_auc of the classifier as a drift_score drift detected when roc_auc > threshold threshold sets the ROC AUC threshold default threshold: 0.55 roc_auc values can be 0 to 1 (typically 0.5 to 1); a higher value means more confident drift detection

Text descriptors drift

You can also check for distribution drift in text descriptors (such as text length, etc.)

To use this method, call a separate TextDescriptorsDriftMetric(). You can pass any of the tabular drift detection methods as a parameter.

report = Report(metrics=[
    TextDescriptorsDriftMetric("Review_Text"),
])

report.run(reference_data=reviews_ref, current_data=reviews_cur, column_mapping=column_mapping)
report