validmind 2.5.25__py3-none-any.whl → 2.6.7__py3-none-any.whl

validmind/tests/model_validation/sklearn/MinimumF1Score.py

@@ -2,24 +2,18 @@
 # See the LICENSE file in the root of this repository for details.
 # SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial
 
-from dataclasses import dataclass
-from typing import List
-
-import pandas as pd
-from numpy import unique
-from sklearn import metrics
-
-from validmind.vm_models import (
-    ResultSummary,
-    ResultTable,
-    ResultTableMetadata,
-    ThresholdTest,
-    ThresholdTestResult,
-)
+import numpy as np
+from sklearn.metrics import f1_score
+
+from validmind.tests import tags, tasks
+from validmind.vm_models import VMDataset, VMModel
 
 
-@dataclass
-class MinimumF1Score(ThresholdTest):
+@tags(
+    "sklearn", "binary_classification", "multiclass_classification", "model_performance"
+)
+@tasks("classification", "text_classification")
+def MinimumF1Score(dataset: VMDataset, model: VMModel, min_threshold: float = 0.5):
     """
     Assesses if the model's F1 score on the validation set meets a predefined minimum threshold, ensuring balanced
     performance between precision and recall.
@@ -59,59 +53,15 @@ class MinimumF1Score(ThresholdTest):
     closely with specific requirements.
     """
 
-    name = "f1_score"
-    required_inputs = ["model", "dataset"]
-    default_params = {"min_threshold": 0.5}
-    tasks = ["classification", "text_classification"]
-    tags = [
-        "sklearn",
-        "binary_classification",
-        "multiclass_classification",
-        "model_performance",
-    ]
-
-    def summary(self, results: List[ThresholdTestResult], all_passed: bool):
-        """
-        The f1 score test returns results like these:
-        [{"values": {"score": 0.734375, "threshold": 0.7}, "passed": true}]
-        """
-        result = results[0]
-        results_table = [
-            {
-                "Score": result.values["score"],
-                "Threshold": result.values["threshold"],
-                "Pass/Fail": "Pass" if result.passed else "Fail",
-            }
-        ]
-
-        return ResultSummary(
-            results=[
-                ResultTable(
-                    data=pd.DataFrame(results_table),
-                    metadata=ResultTableMetadata(title="Minimum F1 Score Test"),
-                )
-            ]
-        )
-
-    def run(self):
-        y_true = self.inputs.dataset.y
-        class_pred = self.inputs.dataset.y_pred(self.inputs.model)
-        y_true = y_true.astype(class_pred.dtype)
-
-        if len(unique(y_true)) > 2:
-            f1_score = metrics.f1_score(y_true, class_pred, average="macro")
-        else:
-            f1_score = metrics.f1_score(y_true, class_pred)
-
-        passed = f1_score > self.params["min_threshold"]
-        results = [
-            ThresholdTestResult(
-                passed=passed,
-                values={
-                    "score": f1_score,
-                    "threshold": self.params["min_threshold"],
-                },
-            )
-        ]
-
-        return self.cache_results(results, passed=all([r.passed for r in results]))
+    if len(np.unique(dataset.y)) > 2:
+        score = f1_score(dataset.y, dataset.y_pred(model), average="macro")
+    else:
+        score = f1_score(dataset.y, dataset.y_pred(model))
+
+    return [
+        {
+            "Score": score,
+            "Threshold": min_threshold,
+            "Pass/Fail": "Pass" if score > min_threshold else "Fail",
+        }
+    ], score > min_threshold

validmind/tests/model_validation/sklearn/MinimumROCAUCScore.py

@@ -2,24 +2,19 @@
 # See the LICENSE file in the root of this repository for details.
 # SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial
 
-from dataclasses import dataclass
-from typing import List
-
 import numpy as np
-import pandas as pd
-from sklearn import metrics, preprocessing
-
-from validmind.vm_models import (
-    ResultSummary,
-    ResultTable,
-    ResultTableMetadata,
-    ThresholdTest,
-    ThresholdTestResult,
-)
+from sklearn.metrics import roc_auc_score
+from sklearn.preprocessing import LabelBinarizer
+
+from validmind.tests import tags, tasks
+from validmind.vm_models import VMDataset, VMModel
 
 
-@dataclass
-class MinimumROCAUCScore(ThresholdTest):
+@tags(
+    "sklearn", "binary_classification", "multiclass_classification", "model_performance"
+)
+@tasks("classification", "text_classification")
+def MinimumROCAUCScore(dataset: VMDataset, model: VMModel, min_threshold: float = 0.5):
     """
     Validates model by checking if the ROC AUC score meets or surpasses a specified threshold.
 
@@ -61,69 +56,25 @@ class MinimumROCAUCScore(ThresholdTest):
     - The use of macro average for multiclass ROC AUC score implies equal weightage to each class, which might not be
     appropriate if the classes are imbalanced.
     """
+    y_true = dataset.y
+
+    if len(np.unique(y_true)) > 2:
+        lb = LabelBinarizer()
+        lb.fit(y_true)
 
-    name = "roc_auc_score"
-    required_inputs = ["model", "dataset"]
-    default_params = {"min_threshold": 0.5}
-    tasks = ["classification", "text_classification"]
-    tags = [
-        "sklearn",
-        "binary_classification",
-        "multiclass_classification",
-        "model_performance",
-    ]
-
-    def summary(self, results: List[ThresholdTestResult], all_passed: bool):
-        """
-        The roc auc score test returns results like these:
-        [{"values": {"score": 0.734375, "threshold": 0.7}, "passed": true}]
-        """
-        result = results[0]
-        results_table = [
-            {
-                "Score": result.values["score"],
-                "Threshold": result.values["threshold"],
-                "Pass/Fail": "Pass" if result.passed else "Fail",
-            }
-        ]
-
-        return ResultSummary(
-            results=[
-                ResultTable(
-                    data=pd.DataFrame(results_table),
-                    metadata=ResultTableMetadata(title="Minimum ROC AUC Score Test"),
-                )
-            ]
+        roc_auc = roc_auc_score(
+            y_true=lb.transform(y_true),
+            y_score=lb.transform(dataset.y_pred(model)),
+            average="macro",
         )
 
-    def multiclass_roc_auc_score(self, y_test, y_pred, average="macro"):
-        lb = preprocessing.LabelBinarizer()
-        lb.fit(y_test)
-        y_test = lb.transform(y_test)
-        y_pred = lb.transform(y_pred)
-        return metrics.roc_auc_score(y_test, y_pred, average=average)
-
-    def run(self):
-        y_true = self.inputs.dataset.y
-
-        if len(np.unique(y_true)) > 2:
-            class_pred = self.inputs.dataset.y_pred(self.inputs.model)
-            y_true = y_true.astype(class_pred.dtype)
-            roc_auc = self.multiclass_roc_auc_score(y_true, class_pred)
-        else:
-            y_prob = self.inputs.dataset.y_prob(self.inputs.model)
-            y_true = y_true.astype(y_prob.dtype).flatten()
-            roc_auc = metrics.roc_auc_score(y_true, y_prob)
-
-        passed = roc_auc > self.params["min_threshold"]
-        results = [
-            ThresholdTestResult(
-                passed=passed,
-                values={
-                    "score": roc_auc,
-                    "threshold": self.params["min_threshold"],
-                },
-            )
-        ]
-
-        return self.cache_results(results, passed=all([r.passed for r in results]))
+    else:
+        roc_auc = roc_auc_score(y_true=y_true, y_score=dataset.y_prob(model))
+
+    return [
+        {
+            "Score": roc_auc,
+            "Threshold": min_threshold,
+            "Pass/Fail": "Pass" if roc_auc > min_threshold else "Fail",
+        }
+    ], roc_auc > min_threshold

validmind/tests/model_validation/sklearn/ModelsPerformanceComparison.py

@@ -2,19 +2,24 @@
 # See the LICENSE file in the root of this repository for details.
 # SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial
 
-from dataclasses import dataclass
-
-from numpy import unique
+import numpy as np
 from sklearn.metrics import classification_report
 
-from validmind.errors import SkipTestError
-from validmind.vm_models import ResultSummary, ResultTable, ResultTableMetadata
+from validmind import tags, tasks
+from validmind.vm_models import VMDataset, VMModel
 
-from .ClassifierPerformance import ClassifierPerformance, multiclass_roc_auc_score
+from .ClassifierPerformance import multiclass_roc_auc_score
 
 
-@dataclass
-class ModelsPerformanceComparison(ClassifierPerformance):
+@tags(
+    "sklearn",
+    "binary_classification",
+    "multiclass_classification",
+    "model_performance",
+    "model_comparison",
+)
+@tasks("classification", "text_classification")
+def ModelsPerformanceComparison(dataset: VMDataset, models: list[VMModel]):
     """
     Evaluates and compares the performance of multiple Machine Learning models using various metrics like accuracy,
     precision, recall, and F1 score.
@@ -57,84 +62,49 @@ class ModelsPerformanceComparison(ClassifierPerformance):
     with unseen data or changes in the data distribution.
     - The ROC AUC score might not be as meaningful or easily interpretable for multilabel/multiclass tasks.
     """
+    y_true = dataset.y
+    classes = {str(i) for i in np.unique(y_true)}
+
+    prf_table = []
+    acc_roc_auc_table = []
+
+    for model in models:
+        y_pred = dataset.y_pred(model)
 
-    name = "models_performance_comparison"
-    required_inputs = ["dataset", "models"]
-    tasks = ["classification", "text_classification"]
-    tags = [
-        "sklearn",
-        "binary_classification",
-        "multiclass_classification",
-        "model_performance",
-        "model_comparison",
-    ]
-
-    def summary(self, metric_value: dict):
-        """
-        This summary varies depending if we're evaluating a binary or multi-class model
-        """
-        results = []
-        prf_table = []
-        classes = {str(i) for i in unique(self.inputs.dataset.y)}
+        report = classification_report(y_true, y_pred, output_dict=True)
+        report["roc_auc"] = multiclass_roc_auc_score(y_true, y_pred)
 
         for class_name in classes:
-            prf_dict = {}
-            prf_dict["Class"] = class_name
-            for m, _ in metric_value.items():
-                prf_dict[f"Precision- {m}"] = metric_value[m][class_name]["precision"]
-                prf_dict[f"Recall- {m}"] = metric_value[m][class_name]["recall"]
-                prf_dict[f"F1- {m}"] = metric_value[m][class_name]["f1-score"]
-            prf_table.append(prf_dict)
-
-        avg_metrics = ["weighted avg", "macro avg"]
-        for class_name in avg_metrics:
-            avg_dict = {}
-            avg_dict["Class"] = class_name
-            for m, _ in metric_value.items():
-                avg_dict[f"Precision- {m}"] = metric_value[m][class_name]["precision"]
-                avg_dict[f"Recall- {m}"] = metric_value[m][class_name]["recall"]
-                avg_dict[f"F1- {m}"] = metric_value[m][class_name]["f1-score"]
-            prf_table.append(avg_dict)
-        results.append(
-            ResultTable(
-                data=prf_table,
-                metadata=ResultTableMetadata(
-                    title="Precision, Recall, and F1 Comparison"
-                ),
+            prf_table.append(
+                {
+                    "Model": model.input_id,
+                    "Class": class_name,
+                    "Precision": report[class_name]["precision"],
+                    "Recall": report[class_name]["recall"],
+                    "F1-Score": report[class_name]["f1-score"],
+                }
             )
-        )
-
-        acc_roc_auc_table = []
-        for metric_name in ["accuracy", "roc_auc"]:
-            acc_roc_auc_dict = {}
-            acc_roc_auc_dict["Metric"] = metric_name
-            for m, _ in metric_value.items():
-                acc_roc_auc_dict[f"accuracy- {m}"] = metric_value[m]["accuracy"]
-                acc_roc_auc_dict[f"roc_auc- {m}"] = metric_value[m]["roc_auc"]
-            acc_roc_auc_table.append(acc_roc_auc_dict)
-        results.append(
-            ResultTable(
-                data=acc_roc_auc_table,
-                metadata=ResultTableMetadata(title="Accuracy and ROC AUC Comparison"),
+        for avg_metric in ["weighted avg", "macro avg"]:
+            prf_table.append(
+                {
+                    "Model": model.input_id,
+                    "Class": avg_metric,
+                    "Precision": report[avg_metric]["precision"],
+                    "Recall": report[avg_metric]["recall"],
+                    "F1-Score": report[avg_metric]["f1-score"],
+                }
             )
-        )
-        return ResultSummary(results=results)
-
-    def run(self):
-        # Check models list is not empty
-        if not self.inputs.models:
-            raise SkipTestError(
-                "List of models must be provided as a `models` parameter to compare performance"
-            )
-
-        all_models = self.inputs.models
 
-        results = {}
-        for idx, model in enumerate(all_models):
-            y_true = self.inputs.dataset.y
-            y_pred = self.inputs.dataset.y_pred(model)
-            report = classification_report(y_true, y_pred, output_dict=True)
-            report["roc_auc"] = multiclass_roc_auc_score(y_true, y_pred)
-            results["model_" + str(idx)] = report
+        for metric in ["accuracy", "roc_auc"]:
+            acc_roc_auc_table.append(
+                {
+                    "Model": model.input_id,
+                    "Metric": metric,
+                    "Value": report[metric],
+                }
+            )
 
-        return self.cache_results(results)
+    return {
+        "Precision, Recall, and F1 Comparison": prf_table,
+        "Accuracy and ROC AUC Comparison": acc_roc_auc_table,
+    }

validmind/tests/model_validation/sklearn/OverfitDiagnosis.py

@@ -2,7 +2,6 @@
 # See the LICENSE file in the root of this repository for details.
 # SPDX-License-Identifier: AGPL-3.0 AND ValidMind Commercial
 
-from dataclasses import dataclass
 from typing import List
 
 import matplotlib.pyplot as plt
@@ -11,17 +10,9 @@ import pandas as pd
 import seaborn as sns
 from sklearn import metrics
 
+from validmind import tags, tasks
 from validmind.logging import get_logger
-from validmind.vm_models import (
-    Figure,
-    ResultSummary,
-    ResultTable,
-    ResultTableMetadata,
-    ThresholdTest,
-    ThresholdTestResult,
-    VMDataset,
-    VMModel,
-)
+from validmind.vm_models import VMDataset, VMModel
 
 logger = get_logger(__name__)
 
@@ -173,56 +164,69 @@ def _plot_overfit_regions(
     return fig
 
 
-# TODO: make this a functional test instead of class-based when appropriate
-# simply have to remove the class and rename this func to OverfitDiagnosis
-def overfit_diagnosis(  # noqa: C901
+@tags(
+    "sklearn",
+    "binary_classification",
+    "multiclass_classification",
+    "linear_regression",
+    "model_diagnosis",
+)
+@tasks("classification", "regression")
+def OverfitDiagnosis(
     model: VMModel,
     datasets: List[VMDataset],
     metric: str = None,
     cut_off_threshold: float = DEFAULT_THRESHOLD,
 ):
-    """Identify overfit regions in a model's predictions.
-
-    This test compares the model's performance on training versus test data, grouped by
-    feature columns. It calculates the difference between the training and test performance
-    for each group and identifies regions where the difference exceeds a specified threshold.
-
-    ## Test Methodology
-
-    This test works for both classification and regression models and with a variety of
-    performance metrics. By default, it uses the AUC metric for classification models and
-    the MSE metric for regression models. The threshold for identifying overfit regions
-    defaults to 0.04 but should be adjusted based on the specific use case.
-
-    ## Inputs
-    - `model` (VMModel): The ValidMind model object to evaluate.
-    - `datasets` (List[VMDataset]): A list of two VMDataset objects where the first dataset
-        is the training data and the second dataset is the test data.
-
-    ## Parameters
-    - `metric` (str, optional): The performance metric to use for evaluation. Choose from:
-        'accuracy', 'auc', 'f1', 'precision', 'recall', 'mse', 'mae', 'r2', 'mape'.
-        Defaults to 'auc' for classification models and 'mse' for regression models.
-    - `cut_off_threshold` (float, optional): The threshold for identifying overfit regions.
-        Defaults to 0.04.
     """
+    Assesses potential overfitting in a model's predictions, identifying regions where performance between training and
+    testing sets deviates significantly.
+
+    ### Purpose
+
+    The Overfit Diagnosis test aims to identify areas in a model's predictions where there is a significant difference
+    in performance between the training and testing sets. This test helps to pinpoint specific regions or feature
+    segments where the model may be overfitting.
+
+    ### Test Mechanism
+
+    This test compares the model's performance on training versus test data, grouped by feature columns. It calculates
+    the difference between the training and test performance for each group and identifies regions where this
+    difference exceeds a specified threshold:
+
+    - The test works for both classification and regression models.
+    - It defaults to using the AUC metric for classification models and the MSE metric for regression models.
+    - The threshold for identifying overfitting regions is set to 0.04 by default.
+    - The test calculates the performance metrics for each feature segment and plots regions where the performance gap
+    exceeds the threshold.
+
+    ### Signs of High Risk
 
-    # Determine if it's a classification or regression model
+    - Significant gaps between training and test performance metrics for specific feature segments.
+    - Multiple regions with performance gaps exceeding the defined threshold.
+    - Higher than expected differences in predicted versus actual values in the test set compared to the training set.
+
+    ### Strengths
+
+    - Identifies specific areas where overfitting occurs.
+    - Supports multiple performance metrics, providing flexibility.
+    - Applicable to both classification and regression models.
+    - Visualization of overfitting segments aids in better understanding and debugging.
+
+    ### Limitations
+
+    - The default threshold may not be suitable for all use cases and requires tuning.
+    - May not capture more subtle forms of overfitting that do not exceed the threshold.
+    - Assumes that the binning of features adequately represents the data segments.
+    """
     is_classification = bool(datasets[0].probability_column(model))
 
-    # Set default metric if not provided
     if not metric:
         metric = (
             DEFAULT_CLASSIFICATION_METRIC
             if is_classification
             else DEFAULT_REGRESSION_METRIC
         )
-        logger.info(
-            f"Using default {'classification' if is_classification else 'regression'} metric: {metric}"
-        )
-
-    if id(cut_off_threshold) == id(DEFAULT_THRESHOLD):
-        logger.info("Using default cut-off threshold of 0.04")
 
     train_df = datasets[0].df
     test_df = datasets[1].df
@@ -279,18 +283,8 @@ def overfit_diagnosis(  # noqa: C901
             )
 
         results = _prepare_results(results_train, results_test, metric)
-
-        fig = _plot_overfit_regions(results, feature_column, cut_off_threshold, metric)
         test_figures.append(
-            Figure(
-                key=f"overfit_diagnosis:{metric}:{feature_column}",
-                figure=fig,
-                metadata={
-                    "metric": metric,
-                    "cut_off_threshold": cut_off_threshold,
-                    "feature": feature_column,
-                },
-            )
+            _plot_overfit_regions(results, feature_column, cut_off_threshold, metric)
         )
 
         for _, row in results[results["gap"] > cut_off_threshold].iterrows():
@@ -306,91 +300,3 @@ def overfit_diagnosis(  # noqa: C901
             )
 
     return {"Overfit Diagnosis": test_results}, *test_figures
-
-
-@dataclass
-class OverfitDiagnosis(ThresholdTest):
-    """
-    Assesses potential overfitting in a model's predictions, identifying regions where performance between training and
-    testing sets deviates significantly.
-
-    ### Purpose
-
-    The Overfit Diagnosis test aims to identify areas in a model's predictions where there is a significant difference
-    in performance between the training and testing sets. This test helps to pinpoint specific regions or feature
-    segments where the model may be overfitting.
-
-    ### Test Mechanism
-
-    This test compares the model's performance on training versus test data, grouped by feature columns. It calculates
-    the difference between the training and test performance for each group and identifies regions where this
-    difference exceeds a specified threshold:
-
-    - The test works for both classification and regression models.
-    - It defaults to using the AUC metric for classification models and the MSE metric for regression models.
-    - The threshold for identifying overfitting regions is set to 0.04 by default.
-    - The test calculates the performance metrics for each feature segment and plots regions where the performance gap
-    exceeds the threshold.
-
-    ### Signs of High Risk
-
-    - Significant gaps between training and test performance metrics for specific feature segments.
-    - Multiple regions with performance gaps exceeding the defined threshold.
-    - Higher than expected differences in predicted versus actual values in the test set compared to the training set.
-
-    ### Strengths
-
-    - Identifies specific areas where overfitting occurs.
-    - Supports multiple performance metrics, providing flexibility.
-    - Applicable to both classification and regression models.
-    - Visualization of overfitting segments aids in better understanding and debugging.
-
-    ### Limitations
-
-    - The default threshold may not be suitable for all use cases and requires tuning.
-    - May not capture more subtle forms of overfitting that do not exceed the threshold.
-    - Assumes that the binning of features adequately represents the data segments.
-    """
-
-    required_inputs = ["model", "datasets"]
-    default_params = {"metric": None, "cut_off_threshold": DEFAULT_THRESHOLD}
-    tasks = ["classification", "regression"]
-    tags = [
-        "sklearn",
-        "binary_classification",
-        "multiclass_classification",
-        "linear_regression",
-        "model_diagnosis",
-    ]
-
-    def run(self):
-        func_result = overfit_diagnosis(
-            self.inputs.model,
-            self.inputs.datasets,
-            metric=self.params["metric"],
-            cut_off_threshold=self.params["cut_off_threshold"],
-        )
-
-        return self.cache_results(
-            test_results_list=[
-                ThresholdTestResult(
-                    test_name=self.params["metric"],
-                    column=row["Feature"],
-                    passed=False,
-                    values={k: v for k, v in row.items()},
-                )
-                for row in func_result[0]["Overfit Diagnosis"]
-            ],
-            passed=(not func_result[0]["Overfit Diagnosis"]),
-            figures=func_result[1:],
-        )
-
-    def summary(self, results, _):
-        return ResultSummary(
-            results=[
-                ResultTable(
-                    data=[result.values for result in results],
-                    metadata=ResultTableMetadata(title="Overfit Diagnosis"),
-                )
-            ],
-        )