chemotools 0.1.9__py3-none-any.whl → 0.1.11__py3-none-any.whl

chemotools/outliers/_base.py

@@ -0,0 +1,188 @@
+from abc import ABC, abstractmethod
+from typing import Optional, Tuple, Union
+
+import numpy as np
+
+from sklearn.base import BaseEstimator, OutlierMixin
+from sklearn.decomposition._base import _BasePCA
+from sklearn.cross_decomposition._pls import _PLS
+from sklearn.pipeline import Pipeline
+from sklearn.utils.validation import check_is_fitted
+
+
+ModelTypes = Union[_BasePCA, _PLS]
+
+
+class _ModelResidualsBase(ABC, BaseEstimator, OutlierMixin):
+    """Base class for model outlier calculations.
+
+    Implements statistical calculations for outlier detection in dimensionality
+    reduction models like PCA and PLS.
+
+    Parameters
+    ----------
+    model : Union[ModelTypes, Pipeline]
+        A fitted _BasePCA or _PLS models or Pipeline ending with such a model
+    confidence : float
+        Confidence level for statistical calculations (between 0 and 1)
+
+    Attributes
+    ----------
+    estimator_ : ModelTypes
+        The fitted model of type _BasePCA or _PLS
+
+    transformer_ : Optional[Pipeline]
+        Preprocessing steps before the model
+
+    n_features_in_ : int
+        Number of features in the input data
+
+    n_components_ : int
+        Number of components in the model
+
+    n_samples_ : int
+        Number of samples used to train the model
+
+    critical_value_ : float
+        The calculated critical value for outlier detection
+    """
+
+    def __init__(
+        self,
+        model: Union[ModelTypes, Pipeline],
+        confidence: float,
+    ) -> None:
+        (
+            self.estimator_,
+            self.transformer_,
+            self.n_features_in_,
+            self.n_components_,
+            self.n_samples_,
+        ) = _validate_and_extract_model(model)
+        self.confidence = _validate_confidence(confidence)
+
+    def fit_predict_residuals(
+        self, X: np.ndarray, y: Optional[np.ndarray] = None
+    ) -> np.ndarray:
+        """Fit the model to the input data and calculate the residuals.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Input data
+
+        y : array-like of shape (n_samples,), default=None
+            Target values
+
+        Returns
+        -------
+        ndarray of shape (n_samples,)
+            The residuals of the model
+        """
+        self.fit(X, y)
+        return self.predict_residuals(X, y, validate=True)
+
+    @abstractmethod
+    def predict_residuals(
+        self, X: np.ndarray, y: Optional[np.ndarray], validate: bool
+    ) -> np.ndarray:
+        """Calculate the residuals of the model.
+
+        Returns
+        -------
+        ndarray of shape (n_samples,)
+            The residuals of the model
+        """
+
+    @abstractmethod
+    def _calculate_critical_value(self, X: np.ndarray) -> float:
+        """Calculate the critical value for outlier detection.
+
+        Returns
+        -------
+        float
+            The calculated critical value for outlier detection
+        """
+
+
+def _get_model_parameters(model: ModelTypes) -> Tuple[int, int, int]:
+    """
+    Get the number of features, components and samples from a model with PLS or PCA. types.
+
+    Parameters
+    ----------
+    model : ModelType
+        A fitted model of type _BasePCA or _PLS
+
+    Returns
+    -------
+    Tuple[int, int, int]
+        The number of features, components and samples in the model
+    """
+    if isinstance(model, _BasePCA):
+        return model.n_features_in_, model.n_components_, model.n_samples_
+    elif isinstance(model, _PLS):
+        return model.n_features_in_, model.n_components, len(model.x_scores_)
+    else:
+        raise ValueError(
+            "Model not a valid model. Must be of base type _BasePCA or _PLS or a Pipeline ending with one of these types."
+        )
+
+
+def _validate_confidence(confidence: float) -> float:
+    """Validate parameters using sklearn conventions.
+
+    Parameters
+    ----------
+    confidence : float
+        Confidence level for statistical calculations (between 0 and 1)
+
+    Returns
+    -------
+    float
+        The validated confidence level
+
+    Raises
+    ------
+    ValueError
+        If confidence is not between 0 and 1
+    """
+    if not 0 < confidence < 1:
+        raise ValueError("Confidence must be between 0 and 1")
+    return confidence
+
+
+def _validate_and_extract_model(
+    model: Union[ModelTypes, Pipeline],
+) -> Tuple[ModelTypes, Optional[Pipeline], int, int, int]:
+    """Validate and extract the model and preprocessing steps.
+
+    Parameters
+    ----------
+    model : Union[ModelTypes, Pipeline]
+        A fitted PCA/PLS model or Pipeline ending with such a model
+
+    Returns
+    -------
+    Tuple[ModelTypes, Optional[Pipeline]]
+        The extracted model and preprocessing steps
+
+    Raises
+    ------
+    ValueError
+        If the model is not of type _BasePCA or _PLS or a Pipeline ending with one of these types or if the model is not fitted
+    """
+    if isinstance(model, Pipeline):
+        preprocessing = model[:-1]
+        model = model[-1]
+    else:
+        preprocessing = None
+
+    if not isinstance(model, (_BasePCA, _PLS)):
+        raise ValueError(
+            "Model not a valid model. Must be of base type _BasePCA or _PLS or a Pipeline ending with one of these types."
+        )
+
+    check_is_fitted(model)
+    n_features_in, n_components, n_samples = _get_model_parameters(model)
+    return model, preprocessing, n_features_in, n_components, n_samples

chemotools/outliers/dmodx.py

@@ -0,0 +1,164 @@
+from typing import Optional, Union
+import numpy as np
+
+from sklearn.pipeline import Pipeline
+from sklearn.utils.validation import validate_data, check_is_fitted
+from scipy.stats import f as f_distribution
+
+
+from ._base import _ModelResidualsBase, ModelTypes
+from .utils import calculate_residual_spectrum
+
+
+class DModX(_ModelResidualsBase):
+    """Calculate Distance to Model (DModX) statistics.
+
+    DModX measures the distance between an observation and the model plane
+    in the X-space, useful for detecting outliers.
+
+    Parameters
+    ----------
+    model : Union[ModelType, Pipeline]
+        A fitted PCA/PLS model or Pipeline ending with such a model
+
+    confidence : float, default=0.95
+        Confidence level for statistical calculations (between 0 and 1)
+
+    Attributes
+    ----------
+    estimator_ : ModelType
+        The fitted model of type _BasePCA or _PLS
+
+    transformer_ : Optional[Pipeline]
+        Preprocessing steps before the model
+
+    n_features_in_ : int
+        Number of features in the input data
+
+    n_components_ : int
+        Number of components in the model
+
+    n_samples_ : int
+        Number of samples used to train the model
+
+    critical_value_ : float
+        The calculated critical value for outlier detection
+
+    train_spe_: float
+        The training sum of squared errors (SSE) for the model normalized by degrees of freedom
+    """
+
+    def __init__(
+        self,
+        model: Union[ModelTypes, Pipeline],
+        confidence: float = 0.95,
+    ) -> None:
+        model, confidence = model, confidence
+        super().__init__(model, confidence)
+
+    def fit(self, X: np.ndarray, y: Optional[np.ndarray] = None) -> "DModX":
+        """
+        Fit the model to the input data.
+
+        This step calculates the critical value for the outlier detection. In the DmodX method,
+        the critical value is not depend on the input data but on the model parameters.
+        """
+        X = validate_data(
+            self, X, y="no_validation", ensure_2d=True, reset=True, dtype=np.float64
+        )
+
+        # Calculate the critical value
+        self.critical_value_ = self._calculate_critical_value()
+
+        # Calculate the degrees of freedom normalized SPE of the training set
+        residuals = calculate_residual_spectrum(X, self.estimator_)
+        squared_errors = np.sum((residuals) ** 2, axis=1)
+        self.train_spe_ = np.sqrt(
+            squared_errors
+            / (self.n_samples_ - self.n_components_ - 1)
+            * (self.n_features_in_ - self.n_components_)
+        )
+
+        return self
+
+    def predict(self, X: np.ndarray) -> np.ndarray:
+        """Identify outliers in the input data.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Input data
+
+        Returns
+        -------
+        ndarray of shape (n_samples,)
+            Boolean array indicating outliers
+        """
+        # Check the estimator has been fitted
+        check_is_fitted(self, ["critical_value_"])
+
+        # Validate the input data
+        X = validate_data(
+            self, X, y="no_validation", ensure_2d=True, reset=True, dtype=np.float64
+        )
+
+        # Calculate outliers based on the DModX statistics
+        dmodx_values = self.predict_residuals(X, validate=False)
+        return np.where(dmodx_values > self.critical_value_, -1, 1)
+
+    def predict_residuals(
+        self, X: np.ndarray, y: Optional[np.ndarray] = None, validate: bool = True
+    ) -> np.ndarray:
+        """Calculate DModX statistics for input data.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Input data
+
+        validate : bool, default=True
+            Whether to validate the input data
+
+        Returns
+        -------
+        ndarray of shape (n_samples,)
+            DModX statistics for each sample
+        """
+        # Check the estimator has been fitted
+        check_is_fitted(self, ["critical_value_"])
+
+        # Validate the input data
+        if validate:
+            X = validate_data(
+                self, X, y="no_validation", ensure_2d=True, reset=True, dtype=np.float64
+            )
+
+        # Apply preprocessing if available
+        if self.transformer_:
+            X = self.transformer_.transform(X)
+
+        # Calculate the DModX statistics
+        residual = calculate_residual_spectrum(X, self.estimator_)
+        squared_errors = np.sum((residual) ** 2, axis=1)
+
+        return (
+            np.sqrt(squared_errors / (self.n_features_in_ - self.n_components_))
+            / self.train_spe_
+        )
+
+    def _calculate_critical_value(self, X: Optional[np.ndarray] = None) -> float:
+        """Calculate F-distribution based critical value.
+
+        Returns
+        -------
+        float
+            The critical value for outlier detection
+        """
+
+        dof_numerator = self.n_features_in_ - self.n_components_
+        dof_denominator = self.n_features_in_ - self.n_components_ - 1
+
+        upper_control_limit = f_distribution.ppf(
+            self.confidence, dof_numerator, dof_denominator
+        )
+        return np.sqrt(upper_control_limit)

chemotools/outliers/hotelling_t2.py

@@ -0,0 +1,156 @@
+from typing import Optional, Union
+import numpy as np
+
+from sklearn.cross_decomposition._pls import _PLS
+from sklearn.decomposition._base import _BasePCA
+from sklearn.pipeline import Pipeline
+from sklearn.utils.validation import validate_data, check_is_fitted
+from scipy.stats import f as f_distribution
+
+from ._base import _ModelResidualsBase, ModelTypes
+
+
+class HotellingT2(_ModelResidualsBase):
+    """
+    Calculate Hotelling's T-squared statistics for PCA or PLS like models.
+
+    Parameters
+    ----------
+    model : Union[ModelType, Pipeline]
+        A fitted PCA/PLS model or Pipeline ending with such a model
+
+    confidence : float, default=0.95
+        Confidence level for statistical calculations (between 0 and 1)
+
+    Attributes
+    ----------
+    estimator_ : ModelType
+        The fitted model of type _BasePCA or _PLS
+
+    transformer_ : Optional[Pipeline]
+        Preprocessing steps before the model
+
+    n_features_in_ : int
+        Number of features in the input data
+
+    n_components_ : int
+        Number of components in the model
+
+    n_samples_ : int
+        Number of samples used to train the model
+
+    critical_value_ : float
+        The calculated critical value for outlier detection
+
+    References
+    ----------
+    Johan A. Westerhuis, Stephen P. Gurden, Age K. Smilde (2001) Generalized contribution plots in multivariate statistical process
+    monitoring  Chemometrics and Intelligent Laboratory Systems 51 2000 95–114
+    """
+
+    def __init__(
+        self, model: Union[ModelTypes, Pipeline], confidence: float = 0.95
+    ) -> None:
+        self.model, self.confidence = model, confidence
+        super().__init__(model, confidence)
+
+    def fit(self, X: np.ndarray, y: Optional[np.ndarray] = None) -> "HotellingT2":
+        """
+        Fit the model to the input data.
+
+        This step calculates the critical value for the outlier detection. In the DmodX method,
+        the critical value is not depend on the input data but on the model parameters.
+        """
+        X = validate_data(
+            self, X, y="no_validation", ensure_2d=True, reset=True, dtype=np.float64
+        )
+
+        self.critical_value_ = self._calculate_critical_value()
+        return self
+
+    def predict(self, X: np.ndarray) -> np.ndarray:
+        """Identify outliers in the input data.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Input data
+
+        Returns
+        -------
+        ndarray of shape (n_samples,)
+            Boolean array indicating outliers
+        """
+        # Check the estimator has been fitted
+        check_is_fitted(self, ["critical_value_"])
+
+        # Validate the input data
+        X = validate_data(
+            self, X, y="no_validation", ensure_2d=True, reset=True, dtype=np.float64
+        )
+
+        # Calculate the Hotelling's T-squared statistics
+        hotelling_t2_values = self.predict_residuals(X, y=None, validate=False)
+        return np.where(hotelling_t2_values > self.critical_value_, -1, 1)
+
+    def predict_residuals(
+        self, X: np.ndarray, y: Optional[np.ndarray] = None, validate: bool = True
+    ) -> np.ndarray:
+        """Calculate Hotelling's T-squared statistics for input data.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Input data
+
+        Returns
+        -------
+        ndarray of shape (n_samples,)
+            Hotellin's T-squared statistics for each sample
+        """
+        # Check the estimator has been fitted
+        check_is_fitted(self, ["critical_value_"])
+
+        # Validate the input data
+        if validate:
+            X = validate_data(
+                self, X, y="no_validation", ensure_2d=True, reset=True, dtype=np.float64
+            )
+
+        # Apply preprocessing steps
+        if self.transformer_:
+            X = self.transformer_.transform(X)
+
+        # Calculate the Hotelling's T-squared statistics
+        if isinstance(self.estimator_, _BasePCA):
+            # For PCA-like models
+            variances = self.estimator_.explained_variance_
+
+        if isinstance(self.estimator_, _PLS):
+            # For PLS-like models
+            variances = np.var(self.estimator_.x_scores_, axis=0)
+
+        # Equivalent to X @ model.components_.T for _BasePCA and X @ model.x_rotations_ for _PLS
+        X_transformed = self.estimator_.transform(X)
+
+        return np.sum((X_transformed**2) / variances, axis=1)
+
+    def _calculate_critical_value(self, X: Optional[np.ndarray] = None) -> float:
+        """
+        Calculate the critical value for the Hotelling's T-squared statistics.
+
+        Returns
+        -------
+        float
+            The critical value for the Hotelling's T-squared statistics
+        """
+
+        critical_value = f_distribution.ppf(
+            self.confidence, self.n_components_, self.n_samples_ - self.n_components_
+        )
+        return (
+            critical_value
+            * self.n_components_
+            * (self.n_samples_ - 1)
+            / (self.n_samples_ - self.n_components_)
+        )

chemotools/outliers/leverage.py

@@ -0,0 +1,151 @@
+from typing import Optional, Union
+import numpy as np
+
+from sklearn.pipeline import Pipeline
+from sklearn.utils.validation import validate_data, check_is_fitted
+
+
+from ._base import _ModelResidualsBase, ModelTypes
+
+
+class Leverage(_ModelResidualsBase):
+    """
+    Calculate the leverage of the training samples on  the latent space of a PCA or PLS models.
+    This method allows to detect datapoints with high leverage in the model.
+
+    Parameters
+    ----------
+    model : Union[ModelType, Pipeline]
+        A fitted PCA/PLS model or Pipeline ending with such a model
+
+    Attributes
+    ----------
+    estimator_ : ModelType
+        The fitted model of type _BasePCA or _PLS
+
+    transformer_ : Optional[Pipeline]
+        Preprocessing steps before the model
+
+    References
+    ----------
+
+    """
+
+    def __init__(
+        self, model: Union[ModelTypes, Pipeline], confidence: float = 0.95
+    ) -> None:
+        model, confidence = model, confidence
+        super().__init__(model, confidence)
+
+    def fit(self, X: np.ndarray, y: Optional[np.ndarray] = None) -> "Leverage":
+        """
+        Fit the model to the input data.
+
+        Parameters
+
+        """
+        X = validate_data(
+            self, X, y="no_validation", ensure_2d=True, reset=True, dtype=np.float64
+        )
+
+        if self.transformer_:
+            X = self.transformer_.fit_transform(X)
+
+        # Compute the critical threshold
+        self.critical_value_ = self._calculate_critical_value(X)
+
+        return self
+
+    def predict(self, X: np.ndarray, y: Optional[np.ndarray] = None) -> np.ndarray:
+        """Calculate Leverage for training data on the model.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Input data
+
+        Returns
+        -------
+        ndarray of shape (n_samples,)
+            Bool with samples with a leverage above the critical value
+        """
+        # Check the estimator has been fitted
+        check_is_fitted(self, ["critical_value_"])
+
+        # Validate the input data
+        X = validate_data(
+            self, X, y="no_validation", ensure_2d=True, reset=True, dtype=np.float64
+        )
+
+        # Preprocess the data
+        if self.transformer_:
+            X = self.transformer_.transform(X)
+
+        # Calculate outliers based on samples with too high leverage
+        leverage = calculate_leverage(X, self.estimator_)
+        return np.where(leverage > self.critical_value_, -1, 1)
+
+    def predict_residuals(
+        self, X: np.ndarray, y: Optional[np.ndarray] = None, validate: bool = True
+    ) -> np.ndarray:
+        """Calculate the leverage of the samples.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            Input data
+
+        Returns
+        -------
+        np.ndarray
+            Leverage of the samples
+        """
+        # Check the estimator has been fitted
+        check_is_fitted(self, ["critical_value_"])
+
+        # Validate the input data
+        if validate:
+            X = validate_data(self, X, ensure_2d=True, dtype=np.float64)
+
+        # Apply preprocessing if available
+        if self.transformer_:
+            X = self.transformer_.transform(X)
+
+        # Calculate the leverage
+        return calculate_leverage(X, self.estimator_)
+
+    def _calculate_critical_value(self, X: np.ndarray) -> float:
+        """Calculate the critical value for outlier detection using the percentile outlier method."""
+
+        # Calculate the leverage of the samples
+        leverage = calculate_leverage(X, self.estimator_)
+
+        # Calculate the critical value
+        return np.percentile(leverage, self.confidence * 100)
+
+
+def calculate_leverage(X: np.ndarray, model: ModelTypes) -> np.ndarray:
+    """
+    Calculate the leverage of the training samples in a PLS/PCA-like model.
+
+    Parameters
+    ----------
+    model : Union[_BasePCA, _PLS]
+        A fitted PCA/PLS model
+
+    X : np.ndarray
+        Preprocessed input data
+
+    Returns
+    -------
+    np.ndarray
+        Leverage of the samples
+    """
+
+    X_transformed = model.transform(X)
+
+    X_hat = (
+        X_transformed @ np.linalg.inv(X_transformed.T @ X_transformed) @ X_transformed.T
+    )
+
+    return np.diag(X_hat)