How to generate multiple locale synthetic data

[esc28]

Environment details

• SDV version: 1.27 Community version
• Python version: 3.13.7
• Operating System: RedHat 8.10

Problem description

I can generate single locale synthetic data like,

from sdv.single_table import GaussianCopulaSynthesizer
synthesizer=GaussianCopulaSynthesizer(
metadata,
locales=['en_CA']
)
synthesizer.fit(data)
synthetic_data=synthesizer.sample(num_rows=500)
synthetic_data.to_csv('filename.csv')

Is there a way to generate 30% Canadian, 40% US and 30% UK synthetic data and save it to one csv file?

[npatki]

Hi @esc28, at the moment, you can supply multiple different locales in the locales list to ensure coverage from each area you want -- Canada, US, and UK.

from sdv.single_table import GaussianCopulaSynthesizer

synthesizer=GaussianCopulaSynthesizer(
  metadata, 
  locales=['en_CA', 'en_GB', 'en_CA']
)

Right now, SDV Community does not allow you to specify the proportion of each. I am curious though -- if your dataset is 30% Canadian, 40% US, and 30% UK, then are there other columns/values in your dataset the denote which one is which? For example, perhaps a categorical column that denotes each row as either Canada, US or UK? In this case, SDV will be able to learn the proportions of each, and there's possibly a way to post-process the data to get what you need.

Or is this 30%/40%/30% breakdown something you want to create from scratch? If you're able to share a bit more about your project, we will be able to guide you better.

[npatki]

Hi @esc28 I'm closing off this thread, as we've answered the question and it has been inactive for some time.

Important: We are in the process of moving to our new support forum. If you'd like to continue this discussion -- or if you have any new questions -- I'd recommend asking on the forum for faster responses from our team. Thanks.