Better file location handeling for META and RANK

KulikDM · KulikDM · commit 54353fa850bb · 2026-05-02T11:13:52.000+03:00
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
@@ -5,44 +5,15 @@ repos:
       - id: trailing-whitespace
       - id: end-of-file-fixer
       - id: check-yaml
-      - id: double-quote-string-fixer
       - id: requirements-txt-fixer
       - id: name-tests-test
         always_run: true
         args: [--pytest-test-first]
       - id: mixed-line-ending
         args: [--fix=lf]
 
-  - repo: https://github.com/DanielNoord/pydocstringformatter
-    rev: v0.7.5
-    hooks:
-      - id: pydocstringformatter
-        args: [--style=pep257, --style=numpydoc]
-        name: Format docstrings
-
-  - repo: https://github.com/asottile/pyupgrade
-    rev: v3.21.2
-    hooks:
-      - id: pyupgrade
-        args: [--py38-plus]
-        name: Upgrade code
-
-  - repo: https://github.com/hhatto/autopep8
-    rev: v2.3.2
-    hooks:
-      - id: autopep8
-        args: [--in-place]
-        name: Format code style
-
-  - repo: https://github.com/PyCQA/isort
-    rev: 7.0.0
-    hooks:
-      - id: isort
-        args: [-m=3]
-        name: Sort imports
-
   - repo: https://github.com/charliermarsh/ruff-pre-commit
-    rev: v0.15.1
+    rev: v0.15.12
     hooks:
       - id: ruff
         args: [--exit-non-zero-on-fix, --fix, --line-length=180]
diff --git a/pythresh/thresholds/meta.py b/pythresh/thresholds/meta.py
@@ -1,5 +1,4 @@
-import os
-from os.path import dirname as up
+from importlib.resources import files
 
 import joblib
 import numpy as np
@@ -18,56 +17,53 @@
 class META(BaseThresholder):
     r"""META class for Meta-modelling thresholder.
 
-       Use a trained meta-model to evaluate a non-parametric means
-       to threshold scores generated by the decision_scores where outliers
-       are set based on the trained meta-model classifier.
-       See :cite:`zhao2022meta` for details.
-
-       Parameters
-       ----------
+    Use a trained meta-model to evaluate a non-parametric means
+    to threshold scores generated by the decision_scores where outliers
+    are set based on the trained meta-model classifier.
+    See :cite:`zhao2022meta` for details.
 
+    Parameters
+    ----------
        method : {'LIN', 'GNB', 'GNBC', 'GNBM'}, optional (default='GNBM')
-           select
-
-           - 'LIN':  RidgeCV trained linear classifier meta-model on true labels
-           - 'GNB':  Gaussian Naive Bayes trained classifier meta-model on true labels
-           - 'GNBC': Gaussian Naive Bayes trained classifier meta-model on best contamination
-           - 'GNBM': Gaussian Naive Bayes multivariate trained classifier meta-model
+        select
 
-       random_state : int, optional (default=1234)
-            Random seed for the random number generators of the thresholders. Can also
-            be set to None.
+        - 'LIN':  RidgeCV trained linear classifier meta-model on true labels
+        - 'GNB':  Gaussian Naive Bayes trained classifier meta-model on true labels
+        - 'GNBC': Gaussian Naive Bayes trained classifier meta-model on best contamination
+        - 'GNBM': Gaussian Naive Bayes multivariate trained classifier meta-model
 
-       Attributes
-       ----------
+    random_state : int, optional (default=1234)
+         Random seed for the random number generators of the thresholders. Can also
+         be set to None.
 
+    Attributes
+    ----------
        thresh_ : threshold value that separates inliers from outliers
 
-       dscores_ : 1D array of decomposed decision scores
-
-       Notes
-       -----
+    dscores_ : 1D array of decomposed decision scores
 
+    Notes
+    -----
        Meta-modelling is the creation of a model of models. If a dataset
-       that contains only the explanatory variables (X), yet no response
-       variable (y), it can still be predicted by using a meta-model. This
-       is done by modelling datasets with known response variables that
-       are similar to the dataset that is missing the response variable.
-
-       The META thresholder was trained using the ``PyOD`` outlier
-       detection methods ``LODA, QMCD, CD, MCD, GMM, KNN, KDE, PCA, Sampling`` and ``IForest``
-       on the AD benchmark datasets: ``ALOI, annthyroid, breastw, campaign, cardio,
-       Cardiotocography, fault, glass, Hepatitis, Ionosphere, landsat, letter, Lymphography,
-       magic.gamma, mammography, mnist, musk, optdigits, PageBlocks, pendigits, Pima,
-       satellite, satimage-2, shuttle, smtp, SpamBase, speech, Stamps, thyroid, vertebral,
-       vowels, Waveform,  WBC, WDBC, Wilt, wine, WPBC, yeast`` available at
-       `ADBench dataset <https://github.com/Minqi824/ADBench/tree/main/adbench/datasets/Classical>`_.
-       META uses a majority vote of all the trained models to determine the
-       inlier/outlier labels.
-
-       Update: the latest GNBC model was further trained on the ``backdoor, celeba, census,
-       cover, donors, fraud, http, InternetAds,`` and ``skin`` datasets and additionally using
-       the ``AutoEncoder, LUNAR, OCSVM, HBOS, KPCA,`` and ``DIF`` outlier detection methods.
+    that contains only the explanatory variables (X), yet no response
+    variable (y), it can still be predicted by using a meta-model. This
+    is done by modelling datasets with known response variables that
+    are similar to the dataset that is missing the response variable.
+
+    The META thresholder was trained using the ``PyOD`` outlier
+    detection methods ``LODA, QMCD, CD, MCD, GMM, KNN, KDE, PCA, Sampling`` and ``IForest``
+    on the AD benchmark datasets: ``ALOI, annthyroid, breastw, campaign, cardio,
+    Cardiotocography, fault, glass, Hepatitis, Ionosphere, landsat, letter, Lymphography,
+    magic.gamma, mammography, mnist, musk, optdigits, PageBlocks, pendigits, Pima,
+    satellite, satimage-2, shuttle, smtp, SpamBase, speech, Stamps, thyroid, vertebral,
+    vowels, Waveform,  WBC, WDBC, Wilt, wine, WPBC, yeast`` available at
+    `ADBench dataset <https://github.com/Minqi824/ADBench/tree/main/adbench/datasets/Classical>`_.
+    META uses a majority vote of all the trained models to determine the
+    inlier/outlier labels.
+
+    Update: the latest GNBC model was further trained on the ``backdoor, celeba, census,
+    cover, donors, fraud, http, InternetAds,`` and ``skin`` datasets and additionally using
+    the ``AutoEncoder, LUNAR, OCSVM, HBOS, KPCA,`` and ``DIF`` outlier detection methods.
 
     """
 
@@ -98,7 +94,6 @@ def eval(self, decision):
             it should be considered as an outlier according to the
             fitted model. 0 stands for inliers and 1 for outliers.
         """
-
         if self._is_fitted is None:
             self._set_attributes(self._attrs, None)
 
@@ -115,8 +110,8 @@ def eval(self, decision):
 
         contam = []
         counts = len(decision)
-        parent = up(up(__file__))
-        model = joblib.load(os.path.join(parent, 'models', clf))
+        model_path = files('pythresh.models').joinpath(clf)
+        model = joblib.load(model_path)
 
         # Sklearn 1.8.0 API patch
         for e in getattr(model, 'estimators_', {}).values():
@@ -194,7 +189,6 @@ def _patch_ridge(est):
     @njit(fastmath=True, parallel=True)
     def _wrap_around_discrepancy(data, check):
         """Wrap-around Quasi-Monte Carlo discrepancy method."""
-
         n = data.shape[0]
         d = data.shape[1]
         p = check.shape[0]
diff --git a/pythresh/utils/rank.py b/pythresh/utils/rank.py
@@ -1,5 +1,4 @@
-import os
-from os.path import dirname as up
+from importlib.resources import files
 
 import numpy as np
 import scipy.stats as stats
@@ -9,89 +8,81 @@
 from sklearn.preprocessing import MinMaxScaler
 from sklearn.utils import check_array
 
-from pythresh.utils.rank_utility import (
-    BREG_metric,
-    Contam_score,
-    GNB_score,
-    mclain_rao_index
-)
+from pythresh.utils.rank_utility import BREG_metric, Contam_score, GNB_score, mclain_rao_index
 
 
-class RANK():
+class RANK:
     """RANK class for ranking outlier detection and thresholding methods.
 
-       Use the RANK class to rank outlier detection and thresholding methods' capabilities
-       to provide the best matthews correlation with respect to the
-       selected threshold method
-
-       Parameters
-       ----------
+    Use the RANK class to rank outlier detection and thresholding methods' capabilities
+    to provide the best matthews correlation with respect to the
+    selected threshold method
 
+    Parameters
+    ----------
        od_models : {list of pyod.model classes}
 
-       thresh : {pythresh.threshold class, float, int, list of pythresh.threshold classes, list of floats, list of ints}
-
-       method : {'model', 'native'}, optional (default='model')
+    thresh : {pythresh.threshold class, float, int, list of pythresh.threshold classes, list of floats, list of ints}
 
-       weights : list of shape 3, optional (default=None)
-             These weights are applied to the combined rank score. The first
-             is for the cdf rankings, the second for the clust rankings, and
-             the third for the mode rankings. Default applies equal weightings
-             to all proxy-metrics. Only applies when method = 'native'.
+    method : {'model', 'native'}, optional (default='model')
 
-       Attributes
-       ----------
+    weights : list of shape 3, optional (default=None)
+          These weights are applied to the combined rank score. The first
+          is for the cdf rankings, the second for the clust rankings, and
+          the third for the mode rankings. Default applies equal weightings
+          to all proxy-metrics. Only applies when method = 'native'.
 
+    Attributes
+    ----------
        cdf_rank_ : list of tuples shape (2, n_od_models) of cdf based rankings
 
-       clust_rank_ : list of tuples shape (2, n_od_models) of cluster based rankings
-
-       consensus_rank_ : list of tuples shape (2, n_od_models) of consensus based rankings
+    clust_rank_ : list of tuples shape (2, n_od_models) of cluster based rankings
 
-       Notes
-       -----
+    consensus_rank_ : list of tuples shape (2, n_od_models) of consensus based rankings
 
+    Notes
+    -----
        The RANK class ranks the outlier detection methods by evaluating
-       three distinct proxy-metrics. The first proxy-metric looks at the outlier
-       likelihood scores by class and measures the cumulative distribution
-       separation using the the Wasserstein distance, and the Exponential Euclidean
-       Bregman distance. The second proxy-metric looks at the relationship between the
-       fitted features (X) and the evaluated classes (y) using the Calinski-Harabasz scores
-       and between the outlier likihood score and the evaluated classes using the
-       Mclain Rao Index. The third proxy-metric evaluates the class difference for each outlier
-       detection and thresholding method with respect to consensus based metrics of all the evaluated
-       outlier detection class labels. This is done using the mean contamination deviation based on
-       TruncatedSVD decomposed scores and Gaussian Naive-Bayes trained consensus score
-
-       Each proxy-metric is ranked separately and a final ranking is applied
-       using all three proxy-metric to get a single ranked result of each
-       outlier detection and thresholding method using the 'native' method. The model method uses
-       a trained LambdaMART ranking model using all the proxy-metrics as input.
-
-       Please note that the data is standardized using
-       ``from pyod.utils.utility import standardizer`` during this ranking process
-
-       Examples
-       --------
-
-       .. code:: python
-
-            # Import libraries
-            from pyod.models.knn import KNN
-            from pyod.models.iforest import IForest
-            from pyod.models.pca import PCA
-            from pyod.models.mcd import MCD
-            from pyod.models.qmcd import QMCD
-            from pythresh.thresholds.filter import FILTER
-            from pythresh.utils.ranking import RANK
-
-            # Initialize models
-            clfs = [KNN(), IForest(), PCA(), MCD(), QMCD()]
-            thres = FILTER()
-
-            # Get rankings
-            ranker = RANK(clfs, thres)
-            rankings = ranker.eval(X)
+    three distinct proxy-metrics. The first proxy-metric looks at the outlier
+    likelihood scores by class and measures the cumulative distribution
+    separation using the the Wasserstein distance, and the Exponential Euclidean
+    Bregman distance. The second proxy-metric looks at the relationship between the
+    fitted features (X) and the evaluated classes (y) using the Calinski-Harabasz scores
+    and between the outlier likihood score and the evaluated classes using the
+    Mclain Rao Index. The third proxy-metric evaluates the class difference for each outlier
+    detection and thresholding method with respect to consensus based metrics of all the evaluated
+    outlier detection class labels. This is done using the mean contamination deviation based on
+    TruncatedSVD decomposed scores and Gaussian Naive-Bayes trained consensus score
+
+    Each proxy-metric is ranked separately and a final ranking is applied
+    using all three proxy-metric to get a single ranked result of each
+    outlier detection and thresholding method using the 'native' method. The model method uses
+    a trained LambdaMART ranking model using all the proxy-metrics as input.
+
+    Please note that the data is standardized using
+    ``from pyod.utils.utility import standardizer`` during this ranking process
+
+    Examples
+    --------
+
+    .. code:: python
+
+         # Import libraries
+         from pyod.models.knn import KNN
+         from pyod.models.iforest import IForest
+         from pyod.models.pca import PCA
+         from pyod.models.mcd import MCD
+         from pyod.models.qmcd import QMCD
+         from pythresh.thresholds.filter import FILTER
+         from pythresh.utils.ranking import RANK
+
+         # Initialize models
+         clfs = [KNN(), IForest(), PCA(), MCD(), QMCD()]
+         thres = FILTER()
+
+         # Get rankings
+         ranker = RANK(clfs, thres)
+         rankings = ranker.eval(X)
     """
 
     def __init__(self, od_models, thresh, method='model', weights=None):
@@ -119,7 +110,6 @@ def eval(self, X):
             thresholder ranked from best to worst in terms of
             performance
         """
-
         X = check_array(X, ensure_2d=True)
         X = standardizer(X)
 
@@ -193,9 +183,9 @@ def eval(self, X):
 
             # Load trained ranking model
             clf = 'rank_model_XGB.json'
-            parent = up(up(__file__))
+            model_path = files('pythresh.models').joinpath(clf)
             ranker = xgb.XGBRanker()
-            ranker.load_model(os.path.join(parent, 'models', clf))
+            ranker.load_model(model_path)
 
             # Transform data
             scaler = MinMaxScaler()
@@ -217,7 +207,6 @@ def eval(self, X):
 
     def _cdf_metric(self, scores, labels):
         """Calculate CDF based metrics."""
-
         if len(np.unique(labels)) == 1:
             return [-1e6, -1e6]
 
@@ -254,7 +243,6 @@ def _cdf_metric(self, scores, labels):
 
     def _clust_metric(self, X, scores, labels):
         """Calculate clustering based metrics."""
-
         if len(np.unique(labels)) == 1:
             return [-1e6, -1e6]
 
@@ -265,15 +253,13 @@ def _clust_metric(self, X, scores, labels):
 
     def _consensus_metric(self, X, scores, labels, contam):
         """Calculate consensus based metrics."""
-
         gnb = GNB_score(X, labels)
         contam = Contam_score(scores, labels, contam)
 
         return np.vstack([gnb, contam]).T.tolist()
 
     def _equi_rank(self, data, order):
         """Get equally weighted rankings from metrics."""
-
         # Get indexes of best to worst for data
         sortings = []
 
@@ -292,10 +278,9 @@ def _equi_rank(self, data, order):
 
     def _rank_sort(self, sortings, weights):
         """Sort weighted rankings."""
-
         # Get unique index values for ranking
         unique_values = {value for ls in sortings for value in ls}
-        scores = {value: 0 for value in unique_values}
+        scores = dict.fromkeys(unique_values, 0)
 
         # Get equally weighted rank
         for value in unique_values:
