Docs
Python API
Evaluator API
create()

create()

Initialize the evaluator instance and prepare for evaluation.

Syntax 

def create() -> None

Parameters

None

Return Value

None

Description

The create() method is used to initialize the evaluator instance. This method must be called after __init__() and before eval().

This method performs the following operations based on the evaluation method specified during initialization:

  1. Load the corresponding evaluation module
  2. Configure evaluation parameters
  3. Initialize the evaluator
  4. Prepare the evaluation environment

Different evaluation methods perform different initialization operations:

  • Anonymeter: Initialize privacy risk evaluator
  • SDMetrics: Initialize data quality evaluator
  • MLUtility: Initialize machine learning utility evaluator
  • Stats: Initialize statistical evaluator
  • Custom: Load and initialize custom evaluator

Basic Examples 

from petsard import Evaluator

# Initialize privacy risk evaluator
evaluator = Evaluator('anonymeter-singlingout')
evaluator.create()  # Initialize evaluator

# Initialize data quality evaluator
evaluator = Evaluator('sdmetrics-qualityreport')
evaluator.create()  # Initialize evaluator

# Initialize machine learning utility evaluator
evaluator = Evaluator(
    'mlutility',
    task_type='classification',
    target='income'
)
evaluator.create()  # Initialize evaluator

Notes

  • Required Step: This method must be called before eval()
  • Single Call: Each evaluator instance only needs to call create() once
  • Parameter Setting: All evaluation parameters must be set during __init__(), create() does not accept parameters
  • Error Handling: If the evaluation method does not exist or parameters are incorrect, an exception will be raised at this stage
  • Resource Initialization: Some evaluators may load models or allocate resources at this stage
  • Best Practice: Use YAML configuration files rather than direct Python API
eval()