API: modeling.classification

skyulf.modeling.classification

Classification models.

AdaBoostClassifierApplier

Bases: SklearnApplier

AdaBoost Classifier Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class AdaBoostClassifierApplier(SklearnApplier):
    """AdaBoost Classifier Applier."""

AdaBoostClassifierCalculator

Bases: SklearnCalculator

AdaBoost Classifier Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("adaboost_classifier", AdaBoostClassifierApplier)
@node_meta(
    id="adaboost_classifier",
    name="AdaBoost Classifier",
    category="Modeling",
    description="An AdaBoost classifier.",
    params={"n_estimators": 50, "learning_rate": 1.0},
)
class AdaBoostClassifierCalculator(SklearnCalculator):
    """AdaBoost Classifier Calculator."""

    def __init__(self):
        super().__init__(
            model_class=AdaBoostClassifier,
            default_params={
                "n_estimators": 50,
                "learning_rate": 1.0,
                "random_state": 42,
            },
            problem_type="classification",
        )

CalibratedClassifierApplier

Bases: SklearnApplier

Calibrated Classifier Applier (well-calibrated predict_proba).

Source code in skyulf-core/skyulf/modeling/classification.py

class CalibratedClassifierApplier(SklearnApplier):
    """Calibrated Classifier Applier (well-calibrated predict_proba)."""

CalibratedClassifierCalculator

Bases: SklearnCalculator

Calibrated Classifier Calculator with a selectable base estimator.

The frontend sends base_estimator as a string key (e.g. "random_forest"); it is resolved here into a fresh estimator instance before CalibratedClassifierCV is constructed. Defaults to logistic regression for backward compatibility.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("calibrated_classifier", CalibratedClassifierApplier)
@node_meta(
    id="calibrated_classifier",
    name="Calibrated Classifier",
    category="Modeling",
    description=(
        "Wraps a base classifier with CalibratedClassifierCV so predicted "
        "probabilities are well-calibrated (Platt/sigmoid or isotonic)."
    ),
    params={"base_estimator": "logistic_regression", "method": "sigmoid", "cv": 5},
    tags=["requires_scaling"],
)
class CalibratedClassifierCalculator(SklearnCalculator):
    """Calibrated Classifier Calculator with a selectable base estimator.

    The frontend sends ``base_estimator`` as a string key (e.g.
    ``"random_forest"``); it is resolved here into a fresh estimator instance
    before ``CalibratedClassifierCV`` is constructed. Defaults to logistic
    regression for backward compatibility.
    """

    # Map of selectable base estimators → factory. Each must support
    # ``predict_proba`` (or ``decision_function``) so calibration is meaningful.
    BASE_ESTIMATORS: Dict[str, Callable[[], BaseEstimator]] = {
        "logistic_regression": lambda: LogisticRegression(max_iter=1000),
        "random_forest": lambda: RandomForestClassifier(n_estimators=100, random_state=42),
        "gradient_boosting": lambda: GradientBoostingClassifier(random_state=42),
        "decision_tree": lambda: DecisionTreeClassifier(random_state=42),
        "gaussian_nb": lambda: GaussianNB(),
        "svc": lambda: SVC(probability=True, random_state=42),
    }

    def __init__(self):
        super().__init__(
            model_class=CalibratedClassifierCV,
            default_params={
                "estimator": LogisticRegression(max_iter=1000),
                "method": "sigmoid",
                "cv": 5,
            },
            problem_type="classification",
        )

    def fit(
        self,
        X: Any,
        y: Any,
        config: Dict[str, Any],
        progress_callback: Optional[Callable[..., Any]] = None,
        log_callback: Optional[Callable[..., Any]] = None,
        validation_data: Any = None,
    ) -> Any:
        config = self._resolve_base_estimator(config)
        return super().fit(X, y, config, progress_callback, log_callback, validation_data)

    @classmethod
    def _resolve_base_estimator(cls, config: Optional[Dict[str, Any]]) -> Dict[str, Any]:
        """Translate a ``base_estimator`` string key into an estimator instance.

        Supports both the flat config shape and the nested ``{"params": {...}}``
        shape used by the model-training payload. Unknown keys fall back to
        logistic regression with a warning.
        """
        if not config:
            return config or {}
        resolved = dict(config)
        nested = isinstance(resolved.get("params"), dict)
        bucket = dict(resolved["params"]) if nested else resolved
        key = bucket.pop("base_estimator", None)
        if isinstance(key, str):
            factory = cls.BASE_ESTIMATORS.get(key)
            if factory is None:
                logger.warning(
                    "Unknown base_estimator '%s'; falling back to logistic_regression.", key
                )
                factory = cls.BASE_ESTIMATORS["logistic_regression"]
            bucket["estimator"] = factory()
        if nested:
            resolved["params"] = bucket
            return resolved
        return bucket

DecisionTreeClassifierApplier

Bases: SklearnApplier

Decision Tree Classifier Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class DecisionTreeClassifierApplier(SklearnApplier):
    """Decision Tree Classifier Applier."""

DecisionTreeClassifierCalculator

Bases: SklearnCalculator

Decision Tree Classifier Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("decision_tree_classifier", DecisionTreeClassifierApplier)
@node_meta(
    id="decision_tree_classifier",
    name="Decision Tree Classifier",
    category="Modeling",
    description="A non-parametric supervised learning method used for classification.",
    params={"max_depth": None, "min_samples_split": 2, "criterion": "gini"},
)
class DecisionTreeClassifierCalculator(SklearnCalculator):
    """Decision Tree Classifier Calculator."""

    def __init__(self):
        super().__init__(
            model_class=DecisionTreeClassifier,
            default_params={
                "max_depth": None,
                "min_samples_split": 2,
                "criterion": "gini",
                "random_state": 42,
            },
            problem_type="classification",
        )

ExtraTreesClassifierApplier

Bases: SklearnApplier

Extra Trees Classifier Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class ExtraTreesClassifierApplier(SklearnApplier):
    """Extra Trees Classifier Applier."""

ExtraTreesClassifierCalculator

Bases: SklearnCalculator

Extra Trees Classifier Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("extra_trees_classifier", ExtraTreesClassifierApplier)
@node_meta(
    id="extra_trees_classifier",
    name="Extra Trees Classifier",
    category="Modeling",
    description="Extremely randomised trees — faster than Random Forest, often comparably accurate.",
    params={"n_estimators": 100, "max_depth": None, "min_samples_split": 2},
)
class ExtraTreesClassifierCalculator(SklearnCalculator):
    """Extra Trees Classifier Calculator."""

    def __init__(self):
        super().__init__(
            model_class=ExtraTreesClassifier,
            default_params={
                "n_estimators": 100,
                "max_depth": None,
                "min_samples_split": 2,
                "min_samples_leaf": 1,
                "criterion": "gini",
                "bootstrap": False,
                "n_jobs": -1,
                "random_state": 42,
            },
            problem_type="classification",
        )

GaussianNBApplier

Bases: SklearnApplier

Gaussian Naive Bayes Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class GaussianNBApplier(SklearnApplier):
    """Gaussian Naive Bayes Applier."""

GaussianNBCalculator

Bases: SklearnCalculator

Gaussian Naive Bayes Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("gaussian_nb", GaussianNBApplier)
@node_meta(
    id="gaussian_nb",
    name="Gaussian Naive Bayes",
    category="Modeling",
    description="Gaussian Naive Bayes (GaussianNB).",
    params={"var_smoothing": 1e-9},
)
class GaussianNBCalculator(SklearnCalculator):
    """Gaussian Naive Bayes Calculator."""

    def __init__(self):
        super().__init__(
            model_class=GaussianNB,
            default_params={"var_smoothing": 1e-9},
            problem_type="classification",
        )

GradientBoostingClassifierApplier

Bases: SklearnApplier

Gradient Boosting Classifier Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class GradientBoostingClassifierApplier(SklearnApplier):
    """Gradient Boosting Classifier Applier."""

GradientBoostingClassifierCalculator

Bases: SklearnCalculator

Gradient Boosting Classifier Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("gradient_boosting_classifier", GradientBoostingClassifierApplier)
@node_meta(
    id="gradient_boosting_classifier",
    name="Gradient Boosting Classifier",
    category="Modeling",
    description="Gradient Boosting for classification.",
    params={"n_estimators": 100, "learning_rate": 0.1, "max_depth": 3},
)
class GradientBoostingClassifierCalculator(SklearnCalculator):
    """Gradient Boosting Classifier Calculator."""

    def __init__(self):
        super().__init__(
            model_class=GradientBoostingClassifier,
            default_params={
                "n_estimators": 100,
                "learning_rate": 0.1,
                "max_depth": 3,
                "random_state": 42,
            },
            problem_type="classification",
        )

HistGradientBoostingClassifierApplier

Bases: SklearnApplier

HistGradientBoosting Classifier Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class HistGradientBoostingClassifierApplier(SklearnApplier):
    """HistGradientBoosting Classifier Applier."""

HistGradientBoostingClassifierCalculator

Bases: SklearnCalculator

HistGradientBoosting Classifier Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("hist_gradient_boosting_classifier", HistGradientBoostingClassifierApplier)
@node_meta(
    id="hist_gradient_boosting_classifier",
    name="Hist Gradient Boosting Classifier",
    category="Modeling",
    description="Histogram-based gradient boosting — sklearn's fast LightGBM-style implementation.",
    params={"max_iter": 100, "learning_rate": 0.1, "max_leaf_nodes": 31},
)
class HistGradientBoostingClassifierCalculator(SklearnCalculator):
    """HistGradientBoosting Classifier Calculator."""

    def __init__(self):
        super().__init__(
            model_class=HistGradientBoostingClassifier,
            default_params={
                "max_iter": 100,
                "learning_rate": 0.1,
                "max_leaf_nodes": 31,
                "max_depth": None,
                "min_samples_leaf": 20,
                "l2_regularization": 0.0,
                "max_bins": 255,
                "random_state": 42,
            },
            problem_type="classification",
        )

KNeighborsClassifierApplier

Bases: SklearnApplier

K-Neighbors Classifier Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class KNeighborsClassifierApplier(SklearnApplier):
    """K-Neighbors Classifier Applier."""

KNeighborsClassifierCalculator

Bases: SklearnCalculator

K-Neighbors Classifier Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("k_neighbors_classifier", KNeighborsClassifierApplier)
@node_meta(
    id="k_neighbors_classifier",
    name="K-Neighbors Classifier",
    category="Modeling",
    description="Classifier implementing the k-nearest neighbors vote.",
    params={"n_neighbors": 5, "weights": "uniform", "algorithm": "auto"},
    tags=["requires_scaling"],
)
class KNeighborsClassifierCalculator(SklearnCalculator):
    """K-Neighbors Classifier Calculator."""

    def __init__(self):
        super().__init__(
            model_class=KNeighborsClassifier,
            default_params={
                "n_neighbors": 5,
                "weights": "uniform",
                "algorithm": "auto",
                "n_jobs": -1,
            },
            problem_type="classification",
        )

LGBMClassifierApplier

Bases: SklearnApplier

LightGBM Classifier Applier.

LightGBM 4.x sets feature_names_in_ to auto-generated names (Column_0, Column_1...) even when fit with numpy arrays, and the property's deleter is intentionally a no-op (see upstream source). That triggers sklearn's UserWarning: X does not have valid feature names on every predict call. We suppress it locally here so the warning never leaks out of the applier boundary.

Source code in skyulf-core/skyulf/modeling/classification.py

class LGBMClassifierApplier(SklearnApplier):
    """LightGBM Classifier Applier.

    LightGBM 4.x sets ``feature_names_in_`` to auto-generated names
    (``Column_0``, ``Column_1``...) even when fit with numpy arrays, and the
    property's deleter is intentionally a no-op (see upstream source). That
    triggers sklearn's ``UserWarning: X does not have valid feature names``
    on every predict call. We suppress it locally here so the warning never
    leaks out of the applier boundary.
    """

    def predict(self, df, model_artifact):
        import warnings

        with warnings.catch_warnings():
            warnings.filterwarnings("ignore", message=".*valid feature names.*")
            return super().predict(df, model_artifact)

    def predict_proba(self, df, model_artifact):
        import warnings

        with warnings.catch_warnings():
            warnings.filterwarnings("ignore", message=".*valid feature names.*")
            return super().predict_proba(df, model_artifact)

LGBMClassifierCalculator

Bases: SklearnCalculator

LightGBM Classifier Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("lgbm_classifier", LGBMClassifierApplier)
@node_meta(
    id="lgbm_classifier",
    name="LightGBM Classifier",
    category="Modeling",
    description="LightGBM: leaf-wise gradient boosting, fast and memory-efficient with categorical support.",
    params={"n_estimators": 100, "num_leaves": 31, "learning_rate": 0.1},
)
class LGBMClassifierCalculator(SklearnCalculator):
    """LightGBM Classifier Calculator."""

    def __init__(self):
        super().__init__(
            model_class=LGBMClassifier,
            default_params={
                "n_estimators": 100,
                "num_leaves": 31,
                "learning_rate": 0.1,
                "max_depth": -1,
                "min_child_samples": 20,
                "subsample": 1.0,
                "colsample_bytree": 1.0,
                "reg_alpha": 0.0,
                "reg_lambda": 0.0,
                "boosting_type": "gbdt",
                "n_jobs": -1,
                "random_state": 42,
                "verbose": -1,
                "verbosity": -1,
            },
            problem_type="classification",
        )

    def fit(
        self, X, y, config, progress_callback=None, log_callback=None, validation_data=None
    ):
        import warnings

        with warnings.catch_warnings():
            warnings.filterwarnings("ignore", message=".*valid feature names.*")
            return super().fit(
                X,
                y,
                config,
                progress_callback=progress_callback,
                log_callback=log_callback,
                validation_data=validation_data,
            )

LogisticRegressionApplier

Bases: SklearnApplier

Logistic Regression Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class LogisticRegressionApplier(SklearnApplier):
    """Logistic Regression Applier."""

LogisticRegressionCalculator

Bases: SklearnCalculator

Logistic Regression Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("logistic_regression", LogisticRegressionApplier)
@node_meta(
    id="logistic_regression",
    name="Logistic Regression",
    category="Modeling",
    description="Linear model for classification.",
    params={"max_iter": 1000, "solver": "lbfgs", "random_state": 42},
    tags=["requires_scaling"],
)
class LogisticRegressionCalculator(SklearnCalculator):
    """Logistic Regression Calculator."""

    def __init__(self):
        super().__init__(
            model_class=LogisticRegression,
            default_params={
                "max_iter": 1000,
                "solver": "lbfgs",
                "random_state": 42,
            },
            problem_type="classification",
        )

RandomForestClassifierApplier

Bases: SklearnApplier

Random Forest Classifier Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class RandomForestClassifierApplier(SklearnApplier):
    """Random Forest Classifier Applier."""

RandomForestClassifierCalculator

Bases: SklearnCalculator

Random Forest Classifier Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("random_forest_classifier", RandomForestClassifierApplier)
@node_meta(
    id="random_forest_classifier",
    name="Random Forest Classifier",
    category="Modeling",
    description="Ensemble of decision trees.",
    params={"n_estimators": 50, "max_depth": 10, "min_samples_split": 5},
)
class RandomForestClassifierCalculator(SklearnCalculator):
    """Random Forest Classifier Calculator."""

    def __init__(self):
        super().__init__(
            model_class=RandomForestClassifier,
            default_params={
                "n_estimators": 50,
                "max_depth": 10,
                "min_samples_split": 5,
                "min_samples_leaf": 2,
                "n_jobs": -1,
                "random_state": 42,
            },
            problem_type="classification",
        )

SGDClassifierApplier

Bases: SklearnApplier

Stochastic Gradient Descent Classifier Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class SGDClassifierApplier(SklearnApplier):
    """Stochastic Gradient Descent Classifier Applier."""

SGDClassifierCalculator

Bases: SklearnCalculator

SGD Classifier Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("sgd_classifier", SGDClassifierApplier)
@node_meta(
    id="sgd_classifier",
    name="SGD Classifier (text / linear)",
    category="Modeling",
    description=(
        "Linear classifiers (SVM, logistic regression, etc.) with SGD training. "
        "Highly efficient for high-dimensional sparse/dense text representations "
        "and large datasets."
    ),
    params={
        "loss": "log_loss",
        "penalty": "l2",
        "alpha": 0.0001,
        "l1_ratio": 0.15,
        "max_iter": 1000,
        "random_state": 42,
    },
    tags=["text", "nlp", "classification", "linear", "requires_scaling"],
)
class SGDClassifierCalculator(SklearnCalculator):
    """SGD Classifier Calculator."""

    def __init__(self):
        super().__init__(
            model_class=SGDClassifier,
            default_params={
                "loss": "log_loss",
                "penalty": "l2",
                "alpha": 0.0001,
                "l1_ratio": 0.15,
                "max_iter": 1000,
                "random_state": 42,
            },
            problem_type="classification",
        )

SVCApplier

Bases: SklearnApplier

SVC Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class SVCApplier(SklearnApplier):
    """SVC Applier."""

SVCCalculator

Bases: SklearnCalculator

SVC Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("svc", SVCApplier)
@node_meta(
    id="svc",
    name="Support Vector Classifier",
    category="Modeling",
    description="C-Support Vector Classification.",
    params={"C": 1.0, "kernel": "rbf", "gamma": "scale"},
    tags=["requires_scaling"],
)
class SVCCalculator(SklearnCalculator):
    """SVC Calculator."""

    def __init__(self):
        super().__init__(
            model_class=SVC,
            default_params={
                "C": 1.0,
                "kernel": "rbf",
                "gamma": "scale",
                "probability": True,
                "random_state": 42,
            },
            problem_type="classification",
        )

XGBClassifierApplier

Bases: SklearnApplier

XGBoost Classifier Applier.

Source code in skyulf-core/skyulf/modeling/classification.py

class XGBClassifierApplier(SklearnApplier):
    """XGBoost Classifier Applier."""

XGBClassifierCalculator

Bases: SklearnCalculator

XGBoost Classifier Calculator.

Source code in skyulf-core/skyulf/modeling/classification.py

@NodeRegistry.register("xgboost_classifier", XGBClassifierApplier)
@node_meta(
    id="xgboost_classifier",
    name="XGBoost Classifier",
    category="Modeling",
    description="Extreme Gradient Boosting classifier.",
    params={"n_estimators": 100, "max_depth": 6, "learning_rate": 0.3},
)
class XGBClassifierCalculator(SklearnCalculator):
    """XGBoost Classifier Calculator."""

    def __init__(self):
        super().__init__(
            model_class=XGBClassifier,
            default_params={
                "n_estimators": 100,
                "max_depth": 6,
                "learning_rate": 0.3,
                "subsample": 0.8,
                "colsample_bytree": 0.8,
                "n_jobs": -1,
                "random_state": 42,
            },
            problem_type="classification",
        )