create()

Initialize and prepare the evaluator for data description.

Syntax

def create() -> None

Parameters

This method takes no parameters.

Returns

• None
  • This method does not return a value

Description

The create() method initializes the internal evaluator implementation based on the configured mode and parameters. This method must be called before using eval() to perform evaluations.

The initialization process includes:

• Creating the appropriate evaluator class (DataDescriber or Stats)
• Configuring statistical methods based on mode
• Setting up comparison parameters if in compare mode
• Preparing internal data structures for evaluation

Basic Examples

Describe Mode

from petsard.evaluator import Describer

# Initialize describer
describer = Describer(
    method='describe',
    describe_method=['mean', 'median', 'std', 'corr']
)

# Create the evaluator
describer.create()

# Now ready to evaluate data
results = describer.eval(data={'data': df})

Compare Mode

from petsard.evaluator import Describer

# Initialize in compare mode
describer = Describer(
    method='describe',
    mode='compare',
    stats_method=['mean', 'std', 'jsdivergence'],
    compare_method='pct_change'
)

# Create the evaluator
describer.create()

# Now ready to compare datasets
results = describer.eval(data={'ori': ori_df, 'syn': syn_df})

With Custom Parameters

from petsard.evaluator import Describer

# Initialize with custom parameters
describer = Describer(
    method='describe',
    describe_method=['mean', 'std', 'min', 'max', 'corr'],
    percentile=0.95  # Calculate 95th percentile
)

# Create the evaluator
describer.create()

# Evaluate data
results = describer.eval(data={'data': df})

Notes

• Required step: Always call create() before eval()
• One-time initialization: Only needs to be called once per Describer instance
• Mode-specific: The created evaluator depends on the configured mode
• Performance: Creation is lightweight, actual computation happens in eval()
• Error handling: Will raise an error if configuration is invalid
• State management: Creates internal state that persists across multiple eval() calls