Python API

Python API

Core Classes

Class	Module	Purpose
ScompLinkPipeline	scomp_link.core	End-to-end pipeline orchestrator
Preprocessor	scomp_link.preprocessing	Data cleaning (polars backend)
FeatureEngineer	scomp_link.preprocessing	Automated feature engineering
DataQualityReport	scomp_link.preprocessing	Data profiling + HTML report
ModelFactory	scomp_link.models	Decision-tree model selection
RegressorOptimizer	scomp_link.models	Grid search regression optimizer
ClassifierOptimizer	scomp_link.models	Grid search classification optimizer
EnsembleOptimizer	scomp_link.models	Voting/stacking ensembles
OptunaOptimizer	scomp_link.models	Bayesian hyperparameter optimization
HalvingSearchOptimizer	scomp_link.models	Successive halving grid search
EarlyStoppingCV	scomp_link.models	Patience-based iteration search
TimeSeriesForecaster	scomp_link.models	ARIMA/SARIMA/ETS forecasting
AnomalyDetector	scomp_link.models	Multi-method anomaly detection
TimeSeriesAnomalyDetector	scomp_link.models	Time series anomaly detection
Validator	scomp_link.validation	Metrics + HTML report
AdvancedCV	scomp_link.validation	LOOCV, Bootstrap
FairnessMetrics	scomp_link.validation	Bias and fairness evaluation
ShapExplainer	scomp_link.explainability	SHAP explanations
LimeExplainer	scomp_link.explainability	LIME explanations
DriftDetector	scomp_link.monitoring	PSI + KS drift detection
ScompArtifact	scomp_link.persistence	Pipeline serialization (.scomp)

Imports

# All main classes are importable from the top level
from scomp_link import (
    ScompLinkPipeline,
    Preprocessor,
    FeatureEngineer,
    DataQualityReport,
    ModelFactory,
    Validator,
    RegressorOptimizer,
    ClassifierOptimizer,
    AnomalyDetector,
    TimeSeriesAnomalyDetector,
    TimeSeriesForecaster,
    FairnessMetrics,
    ShapExplainer,
    LimeExplainer,
    DriftDetector,
    ScompArtifact,
    set_verbosity,
)

# Advanced tuning (not top-level)
from scomp_link.models.advanced_tuning import OptunaOptimizer, HalvingSearchOptimizer, EarlyStoppingCV

Logging Control

import scomp_link

scomp_link.set_verbosity("silent")   # no output at all
scomp_link.set_verbosity("warning")  # only warnings
scomp_link.set_verbosity("info")     # default (progress messages)
scomp_link.set_verbosity("debug")    # verbose debugging

External code can also access the logger directly:

import logging
logging.getLogger("scomp_link").setLevel(logging.WARNING)