Quick start: Synthetic data

Synthetic data is artificially generated data that mimics the structure and statistical properties of real data, but contains no actual personal or sensitive information.

Use a generator to create sythetic data based on your requirements. Follow the generator quick start guide if you want to create your own generator.

Step 1: Generate synthetic data

Navigate to the generator that you wish to use by clicking Generators in the left-side navigation menu and selecting from the available generators.
On the generator page, click Generate data.
Configure the generation parameters based on your requirements:

Parameter	Description
Sample size	Define the number of rows to generate for your synthetic data.
Conditional simulation	Use a seed dataset as context for conditional simulation.
Sampling controls	Adjust how closely your synthetic data mirrors the original distribution.
Fairness	Apply statistical parity to a target column with respect to selected sensitive columns, ensuring statistical independence.
Rebalancing	Rebalance the distribution in a categorical column based on specified categories and their new percentages.
Imputation	Replace missing values in a column with statistically coherent values from the same column.
Data report	Control data report generation for this model. Disable it to speed up the process of finalizing synthetic dataset generation.

Click Start generation.

Follow progress in the Generation status section on the synthetic dataset page.

What’s next

When the generation of synthetic data completes, you can download the data in CSV, Parquet, or XSLX format.

You can also make the synthetic dataset public so others can access its page, review the Data report, Data samples, Generation steps, and Configuration, or download the data in their preferred format.

Start a new chat with the Assistant by clicking New chat in the left-side navigation menu.
Prompt the Assistant to connect to the synthetic data generator that you would like to use. Alternatively, you can click Explore from the generator page of any generator.

Connect to the Baseball Generator and describe this resource.

Define the synthetic dataset that you would like to generate with this generator.

Generate a synthetic dataset containing 1000 players with a batting average over 0.300 using this generator.

💡

The Assitant can help you generate code to use with the Synthetic Data SDK as well!

Describe your use case and the resources that you’d like to use and ask the Assistant to provide the code to use with the Synthetic Data SDK.

Available in both Local and Client modes.

Use the imported generator object g to generate a new synthetic dataset.

Generate your first synthetic dataset, using the imported generator.

python

# Local mode
sd = mostly.generate(g, size=100)
 
# Client mode
# sd = mostly_remote.generate(g, size=100)

Consume the generated data. python copy filename="python" df = sd.data() df

Step 2: Download synthetic data

With the synthetic dataset open, click Download synthetic dataset.
Select to download as CSV, Parquet, or Excel.

Step 3: Make the synthetic dataset public

With the synthetic dataset open, click Share.
Select Public from the Visibility dropdown
Click Save.

Now you can copy and share the URL of the synthetic dataset.

What’s next

Use the synthetic dataset to freely analyze the data without the privacy-related concerns.

Share the dataset URL so others can do the same.