synthetic-eval 0.1.0

Package for Evaluation of Synthetic Tabular Data Quality

Synthetic-Eval

Synthetic-Eval is a package for the comprehensive evaluation of synthetic tabular datasets.

1. Installation

Install using pip:

pip install synthetic-eval

2. Supported Metrics

• Statistical Fidelity
  1. KL-Divergence (KL)
  2. Goodness-of-Fit (Kolmogorov-Smirnov test & Chi-Squared test) (GoF)
  3. Maximum Mean Discrepancy (MMD)
  4. Cramer-Wold Distance (CW)
  5. (naive) $\alpha$-precision & $\beta$-recall (alpha_precision, beta_recall)
• Machine Learning Utility (classification task)
  1. Accuracy (base_cls, syn_cls)
  2. Model Selection Performance (model_selection)
  3. Feature Selection Performance (feature_selection)
• Privacy Preservation
  1. $k$-Anonymization (Kanon_base, Kanon_syn)
  2. $k$-Map (KMap)
  3. Distance to Closest Record (DCR_RS, DCR_RR, DCR_SS)
  4. Attribute Disclosure (AD)

3. Usage

from synthetic_eval import evaluation
evaluation.evaluate # function for evaluating synthetic data quality

• See example.ipynb for detailed example and its results with loan dataset.
  • Link for download loan dataset: https://www.kaggle.com/datasets/teertha/personal-loan-modeling

Example

• Please ensure that the target column for the machine learning utility is the last column of the dataset.

"""import libraries"""
import pandas as pd
import torch
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

"""specify column types"""
data = pd.read_csv('./loan.csv') 
# len(data) # 5,000

"""specify column types"""
data = pd.read_csv('./loan.csv') 
# len(data) # 5,000

"""specify column types"""
continuous_features = [
    'Age',
    'Experience',
    'Income', 
    'CCAvg',
    'Mortgage',
]
categorical_features = [
    'Family',
    'Securities Account',
    'CD Account',
    'Online',
    'CreditCard',
    'Personal Loan', 
]
target = 'Personal Loan' # machine learning utility target column

### the target column should be the last column
data = data[continuous_features + [x for x in categorical_features if x != target] + [target]] 

"""training, test, synthetic datasets"""
data[categorical_features] = data[categorical_features].apply(
        lambda col: col.astype('category').cat.codes) 

train = data.iloc[:2000]
test = data.iloc[2000:4000]
syndata = data.iloc[4000:]

"""load Synthetic-Eval"""
from synthetic_eval import evaluation
results = evaluation.evaluate(
    syndata, train, test, 
    target, continuous_features, categorical_features, device
)

"""print results"""
for x, y in results._asdict().items():
    print(f"{x}: {y:.3f}")

3. References

• https://github.com/vanderschaarlab/synthcity/blob/main/src/synthcity/metrics
• https://github.com/HLasse/data-centric-synthetic-data