Synthesizer

Synthesizer(
    method,
    **kwargs
)

Synthetic data generator supporting multiple synthesis methods.

Parameters

method (str): Synthesis method
- ‘default’: Use SDV-GaussianCopula
- ‘custom_data’: Load custom data from file
- ‘sdv-single_table-{method}’: Use SDV provided methods
  - copulagan: CopulaGAN generative model
  - ctgan: CTGAN generative model
  - gaussiancopula: Gaussian Copula model
  - tvae: TVAE generative model

Examples

from petsard import Synthesizer


# Using SDV's GaussianCopula
syn = Synthesizer(method='sdv-single_table-gaussiancopula')

# Using default method
syn = Synthesizer(method='default')

# Synthesizng
syn.create(data=df, metadata=metadata)
syn.fit_sample()
synthetic_data = syn.data_syn

Methods

`create()`

syn.create(data, metadata=None)

Initialize synthesizer.

Parameters

data (pd.DataFrame): Training dataset
metadata (Metadata, optional): Dataset’s Metadata object
- Default: None

Returns

None. Initializes the synthesizer object

`fit()`

Train synthesis model.

Parameters

None

Returns

None. Updates synthesizer’s internal state

`sample()`

syn.sample(
    sample_num_rows=None,
    reset_sampling=False,
    output_file_path=None
)

Generate synthetic data.

Parameters

sample_num_rows (int, optional): Number of rows to generate
- Default: None (use original data row count)
reset_sampling (bool, optional): Whether to reset sampling state
- Default: False
output_file_path (str, optional): Output file path
- Default: None

Returns

None. Generated data is stored in data_syn attribute

`fit_sample()`

syn.fit_sample(**kwargs)

依序執行訓練與生成。整合 fit() 和 sample() 的功能。

Parameters

Same as sample()

Returns

None. Generated data is stored in data_syn attribute

Attributes

data_syn: Generated synthetic data (pd.DataFrame)
config: Configuration dictionary containing:
- method (str): Synthesis method name
- method_code (int): Method type code
- Additional parameters specific to each method
synthesizer: Instantiated synthesizer object (for SDV methods)
loader: Loader object (for ‘custom_data’ method only)

Processor Constrainer