Categorical Distribution
Categorical reference statistics between two datasets.
Validator Overview
The categorical distribution Validator verifies that only the expected number of categories are added, removed, or changed over time, as well as the relative entropy.
| Metric Options | Description |
|---|---|
| Categories added | Validates the number of new categories in the source dataset against a reference dataset. |
| Categories removed | Validates the number of missing categories source dataset against a reference dataset. |
| Categories changed | Validates the number of new and removed categories in the source dataset against a reference dataset. |
| Relative entropy | Validates distribution shifts in your data over time. |
Relative Entropy
You can use relative entropy to validate distribution shifts in your data over time, or to compare the distributions of two data sets. Relative entropy is presented as a percentage where:
- 0% means identical empirical distributions.
- 100% means maximal difference in empirical distributions.
Note
In Validio, relative entropy is adapted from the implementation of the Kullback - Leibler divergence.
Metric Configuration Parameters
The following parameters are used in the Metric configuration step of creating a Categorical Distribution validator.
| Parameter | Description | Options |
|---|---|---|
Metric |
Select the metric to calculate. |
Categories Added |
Field |
Select a source field to use for the calculation. |
List of available fields with data type string. |
Reference Field |
Select a reference source field to use for the calculation. |
List of available fields with data type string. |
Filter |
(Optional) Use filters to specify which records to include in the calculation. |
List of existing filters or create a new filter. |
Reference Filter |
(Optional) Use filters to specify which reference records to include in the calculation. |
List of existing filters or create a new filter. |
Window |
Use windows to define the time-range over which the data is aggregated. |
List of existing windows or create a new window. |
Reference Window Offset |
The number of windows you want to offset the aggregation. |
Enter a number. |
Number of Reference Windows |
The number of windows to include. |
Enter a number. |
Segmentation |
Use segmentation to break the data into separate groups for analysis. |
List of existing segmentations, Unsegmented (default), or create a new segmentation. |
Initialize using historic data |
Start the validator with historical data to prime the anomaly detection algorithms. |
Metric Calculation Example
The following example illustrates how the categorical distribution validator calculates the different metrics. The table shows all values from the categorical fields monitored in respective datasets:
| Categories in the reference dataset | Categories in the source dataset |
|---|---|
| A | |
| B | |
| C | C |
| D | D |
| E | E |
| F |
| Metric | Example Result |
|---|---|
| Categories added | In the example, compared to the reference dataset, the source dataset has one new categorical value F. The number of new categories is 1. |
| Categories removed | In the example, two categorical values are missing in the source dataset vs. reference dataset; A and B. The number of removed categories is 2. |
| Categories changed | In the example, the number of changed categories is the sum of new and removed categories. In this case, 1+2=3. |
Updated 11 days ago