Skip to content

advanced_profiler

sqllocks_spindle.inference.advanced_profiler

Advanced fidelity profiling — Tier 1 improvements.

Adds multi-variate conditional profiling, GMM distribution fitting, adversarial distinguishability scoring, temporal/sequence profiling, and FFT periodicity detection on top of the base DataProfiler.

All features degrade gracefully when optional dependencies are absent: - sklearn: GMM fitting + adversarial scoring - scipy: periodicity (FFT) + KS test for sequence gaps

Classes

GMMFit dataclass

Gaussian Mixture Model fit for a numeric column.

ConditionalProfile dataclass

Conditional statistics for col_a given values of col_b.

AdversarialResult dataclass

Result of the adversarial (distinguishability) test.

Attributes
distinguishability_score property

0 = perfectly indistinguishable, 100 = perfectly distinguishable.

TemporalProfile dataclass

Temporal / sequence analysis for a datetime or sorted numeric column.

PeriodicityResult dataclass

FFT-based periodicity detection result.

AdvancedTableProfile dataclass

Extended profile combining base stats with Tier 1 fidelity features.

AdvancedProfiler

Runs Tier 1 fidelity profiling on a pair of DataFrames (real + synthetic).

Usage::

profiler = AdvancedProfiler()
adv = profiler.profile_pair(real_df, synth_df, table_name="orders")
print(f"AUC: {adv.adversarial.auc_roc:.3f}")
Methods:
profile_pair(real, synthetic, table_name='table')

Profile real + synthetic DataFrames and return AdvancedTableProfile.

profile_single(df, table_name='table')

Profile a single DataFrame (no adversarial test — needs both real+synth).