Postprocessor YAML

YAML configuration file format for the Postprocessor module, used for data postprocessing (inverse transformation).

Usage Example

Click the button below to run the example in Colab:

Loader:
  load_benchmark_with_schema:
    filepath: benchmark://adult-income
    schema: benchmark://adult-income_schema
Preprocessor:
  default:
    method: default
Synthesizer:
  default:
    method: default
Postprocessor:
  default:
    method: default  # Postprocessing: automatically restore preprocessing transformations
Reporter:
  save_all_step:
    method: save_data
    source:
      - Preprocessor
      - Synthesizer
      - Postprocessor

Main Parameters

• method (string, required)
  • Currently only the 'default' method is available, which automatically detects the corresponding Preprocessor configuration and performs inverse restoration operations

How It Works

The Postprocessor automatically performs inverse operations of the Preprocessor, restoring synthetic data to its original format:

1. Auto-detection: Reads the corresponding Preprocessor configuration
2. Reverse execution: Executes restoration operations in reverse order
3. Data restoration:
   • Inverse scaling (scaler → inverse scaler)
   • Inverse encoding (encoder → inverse encoder)
   • Inverse discretization (discretizing → inverse discretizing)
   • Restore missing values (missing → restore NA)

Mapping with Preprocessor

Restoration Sequence

Assuming the Preprocessor sequence is ['missing', 'outlier', 'encoder', 'scaler'], the Postprocessor restoration sequence will be ['scaler', 'encoder', 'missing']

Note:

• outlier will not be restored (outlier handling cannot be reversed)
• Sequence is automatically reversed
• Missing values are re-inserted according to the original data proportions

Restoration Flowchart

Synthetic Data (encoded)
  ↓
Inverse Scaling (Scaler inverse)
  ↓
Inverse Encoding (Encoder inverse)
  ↓
Insert Missing Values (Missing restore)
  ↓
Original Format Data

Precision Restoration

Postprocessor automatically restores numeric fields to their original precision:

• Precision Source: Uses the schema from Preprocessor input (preprocessor_input_schema)
• Auto-Application: Data is rounded according to original precision after restoration
• Precision Consistency: Ensures restored data matches original data precision

Execution Instructions

• Postprocessor must be executed after Preprocessor
• The system will automatically read the previous Preprocessor configuration
• The system will automatically retrieve original data precision information
• If there is no corresponding Preprocessor, the Postprocessor will not perform any operations

Notes

• Postprocessor automatically detects and uses the corresponding Preprocessor configuration
• Outlier handling cannot be restored, so data ranges may differ slightly from the original data
• Missing values are randomly inserted according to the original proportions (positions may differ)
• Data types are automatically aligned with the original schema definition
• Precision is automatically restored to match original data precision
• Multiple Preprocessor/Postprocessor Configuration Recommendations:
  • Even with multiple Preprocessors configured, typically only a single Postprocessor is needed to automatically handle all restoration operations
  • Testing for multiple Preprocessor combinations and multiple Postprocessor configurations is currently not comprehensive
  • Recommendation: Avoid trying multiple complex combinations of Preprocessors simultaneously to ensure postprocessing stability
• For detailed restoration mechanisms, please refer to the Postprocessor API documentation