Ctrlk
ForumDataCebo
For the complete documentation index, see llms.txt. This page is also available as Markdown.

⭐Quality Report API

The SDMetrics Quality Report captures the statistical similarity between real data and synthetic data. If the synthetic and real data are statistically similar, we refer to the synthetic data as being high quality (aka high fidelity). The report runs select metrics to measure these properties and summarizes the results.

from sdmetrics.reports import QualityReport

report = QualityReport()
report.generate(real_data, synthetic_data, metadata)
Generating report ...

(1/4) Evaluating Column Shapes: |██████████| 15/15 [00:00<00:00, 564.15it/s]|
Column Shapes Score: 85.61%

(2/4) Evaluating Column Pair Trends: |██████████| 55/55 [00:00<00:00, 110.40it/s]|
Column Pair Trends Score: 71.97%

(3/4) Evaluating Cardinality: |██████████| 1/1 [00:00<00:00, 53.27it/s]|
Cardinality Score: 70.0%

(4/4) Evaluating Intertable Trends: |██████████| 50/50 [00:00<00:00, 86.54it/s]|
Intertable Trends Score: 68.49%

Overall Score (Average): 74.02%

Once you have generated the report you can get more details to explain the results, visualize the scores and save the report to share it.

Usage

Generating the report

QualityReport()

Create your report object by importing it from the multi table reports module.

generate(real_data, synthetic_data, metadata)

Generate your report by passing in the data and metadata.

  • (required) real_data: A dictionary mapping the name of each table to a pandas.DataFrame containing the real data for that table

  • (required) synthetic_data: A dictionary mapping the name of each table to a pandas.DataFrame containing the synthetic data for that table

  • (required) metadata: A dictionary describing the format, types of data and relationship between the tables. See Multi Table Metadata for more details.

  • verbose: A boolean describing whether or not to print the report progress and results. Defaults to True. Set this to False to run the report silently.

Once completed, some preliminary scores will be printed out.

Getting & explaining the results

Every score that the report generates ranges from 0 (worst) to 1 (best)

get_score()

Use this method at any point to retrieve the overall score.

Returns: A floating point value between 0 and 1 that summarizes the quality of your synthetic data.

get_properties()

Use this method at any point to retrieve each property that the report evaluated

Returns: A pandas.DataFrame that lists each property name and its associated score

get_details(property_name, table_name)

Use this method to get more details about a particular property.

  • (required) property_name: A string with the name of the property. One of: 'Column Shapes', 'Column Pair Trends', 'Cardinality' or 'Intertable Trends'

  • table_name: A string with the name of the table. If provided, you'll receive filtered results for the table.

Returns: A pandas.DataFrame that returns more details about the property for the given table

For example, the details for 'Column Shapes' shows the name of each individual column, the metric that was used to compute it and the overall score for that column.

Visualizing the report

You can visualize the properties and use the SDMetrics utilities to visualize the raw data too.

get_visualization(property_name, table_name)

Use this method to visualize the details about a property.

  • (required) property_name: A string with the name of the property. One of: 'Column Shapes', 'Column Pair Trends', 'Cardinality' or 'Intertable Trends'

  • (required) table_name: A string with the name of the table

Returns: A plotly.Figure object

The exact visualization is based on the property. For example, 'Column Shapes' property visualizes the quality score for every column as well as the metric used to compute it.

Other visualizations are available! Use the SDMetrics Visualization Utilities to get more insights into your data. Tip: All visualizations returned in this report are interactive. If you're using an iPython notebook, you can zoom, pan, toggle legends and take screenshots.

Saving & loading the report

You can save your report if you want to share or access it in the future.

save(filepath)

Save the Python report object

  • (required) filepath: The name of file to save the object. This must end with .pkl

The report does not save the full real and synthetic datasets, but it does save the metadata along with the score for each property, breakdown and metric.

The score information may still leak sensitive details about your real data. Use caution when deciding where to store the report and who to share it with.

QualityReport.load(filepath)

Load the report from the file

  • (required) filepath: The name of the file where the report is stored

Returns: A QualityReport object.

FAQs

What is the best way to see the visualizations? Can I save them?

This report returns all visualizations as plotly.Figure object, which are integrated with most iPython notebooks (eg. Colab, Jupyter).

Tip! You can interact with the visualizations when you're viewing them in a notebook. You can zoom, pan and take screenshots.

It's also possible to programmatically save a static image export. See the Plotly Guide for more details.

Can this report check for similarity in higher orders?

Higher order distributions of 3 or more columns are not included in the Quality Report. We have found that very high order similarity may have an adverse effect on the synthetic data usability; after a certain point, it indicates that the synthetic data is just a copy of the real data. (For more information, see Privacy Metrics.)

If higher order similarity is a requirement, you likely have a targeted use case for synthetic data (eg. machine learning efficacy). Until we add these reports, you may want to explore other metrics.

Is the score deterministic?

Every metric that the quality report computes is deterministic. However, if your dataset contains over 50K rows, the quality report may subsample your data to compute specific metrics such as ContingencySimilarity. This improves the performance.

Since the subsampling is random, you may observe that the score can vary slightly if you re-run the quality report. We've chosen the threshold of 50K such that the variation in the quality score will only be a small percentage off from the actual value.

If you'd like to turn subsampling off and make your score deterministic, you can set the num_rows_subsample attribute to None. (Though keep in mind that this can increase the time it takes to generate the report.)

PreviousQualityNextCardinalityShapeSimilarity

Last updated