few utnittests (#46)

* few utnittests
janoPig · Jan 16, 2024 · c7b7cad · c7b7cad
1 parent 087314b
commit c7b7cad
Show file tree

Hide file tree

Showing 3 changed files with 54 additions and 21 deletions.
diff --git a/HROCH/classifier.py b/HROCH/classifier.py
@@ -280,13 +280,9 @@ class SymbolicClassifier(OneVsRestClassifier):
     ----------
     kwargs : Any
         Parameters passed to [NonlinearLogisticRegressor](https://janopig.github.io/HROCH/HROCH.html#NonlinearLogisticRegressor) estimator
-    
-    verbose : int
-        Verbosity level for OneVsRestClassifier
-
     """
-    def __init__(self, verbose=0, **kwargs):
-        super().__init__(estimator=NonlinearLogisticRegressor(**kwargs), verbose=verbose)  
+    def __init__(self, **kwargs):
+        super().__init__(estimator=NonlinearLogisticRegressor(**kwargs))  
 
     def fit(self, X: numpy.ndarray, y: numpy.ndarray):
         """
@@ -307,7 +303,7 @@ def fit(self, X: numpy.ndarray, y: numpy.ndarray):
             Fitted estimator.
         """
 
-        super(OneVsRestClassifier, self).fit(X, y)
+        super().fit(X, y)
         return self
 
     def predict(self, X: numpy.ndarray):
@@ -324,7 +320,7 @@ def predict(self, X: numpy.ndarray):
         y : numpy.ndarray of shape (n_samples,)
             The predicted classes.
         """
-        return super(OneVsRestClassifier, self).predict(X)
+        return super().predict(X)
 
     def predict_proba(self, X: numpy.ndarray):
         """
@@ -340,4 +336,4 @@ def predict_proba(self, X: numpy.ndarray):
             The class probabilities of the input samples. The order of the
             classes corresponds to that in the attribute :term:`classes_`.
         """
-        return super(OneVsRestClassifier, self).predict_proba(X)
+        return super().predict_proba(X)
diff --git a/HROCH/fuzzy.py b/HROCH/fuzzy.py
@@ -2,6 +2,7 @@
 from .hroch import SymbolicSolver
 from sklearn.base import ClassifierMixin
 from sklearn.metrics import log_loss, make_scorer
+from sklearn.utils import compute_class_weight
 import numpy as numpy
 from typing import Iterable
 
@@ -190,6 +191,8 @@ def fit(self, X: numpy.ndarray, y: numpy.ndarray, sample_weight=None, check_inpu
 
         self.classes_ = numpy.unique(y)
         self.n_classes_ = len(self.classes_)
+
+        self.class_weight_ = compute_class_weight(self.class_weight, classes=self.classes_, y=y)
 
         super(FuzzyRegressor, self).fit(X, y, sample_weight=sample_weight, check_input=check_input)
         return self
@@ -249,14 +252,9 @@ class FuzzyClassifier(OneVsRestClassifier):
     ----------
     kwargs : Any
         Parameters passed to [FuzzyRegressor](https://janopig.github.io/HROCH/HROCH.html#FuzzyRegressor) estimator
-        
-    verbose : int
-        Verbosity level for OneVsRestClassifier
-
     """
-    def __init__(self, verbose=0, **kwargs):
-        super().__init__(estimator=FuzzyRegressor(**kwargs), verbose=verbose)
-
+    def __init__(self, **kwargs):
+        super().__init__(estimator=FuzzyRegressor(**kwargs))
 
     def fit(self, X: numpy.ndarray, y: numpy.ndarray):
         """
@@ -277,7 +275,7 @@ def fit(self, X: numpy.ndarray, y: numpy.ndarray):
             Fitted estimator.
         """
 
-        super(OneVsRestClassifier, self).fit(X, y)
+        super().fit(X, y)
         return self
 
     def predict(self, X: numpy.ndarray):
@@ -294,7 +292,7 @@ def predict(self, X: numpy.ndarray):
         y : numpy.ndarray of shape (n_samples,)
             The predicted classes.
         """
-        return super(OneVsRestClassifier, self).predict(X)
+        return super().predict(X)
 
     def predict_proba(self, X: numpy.ndarray):
         """
@@ -310,4 +308,4 @@ def predict_proba(self, X: numpy.ndarray):
             The class probabilities of the input samples. The order of the
             classes corresponds to that in the attribute :term:`classes_`.
         """
-        return super(OneVsRestClassifier, self).predict_proba(X)
+        return super().predict_proba(X)
diff --git a/test/sklearn_test.py b/test/sklearn_test.py
@@ -1,11 +1,11 @@
 import HROCH
-from HROCH import SymbolicRegressor, NonlinearLogisticRegressor
+from HROCH import SymbolicRegressor, NonlinearLogisticRegressor, SymbolicClassifier, FuzzyRegressor, FuzzyClassifier
 import unittest
 import numpy as np
 import pandas as pd
 from sklearn.datasets import load_breast_cancer
 from sklearn.model_selection import train_test_split
-from sklearn.metrics import r2_score
+from sklearn.metrics import r2_score, log_loss, make_scorer, roc_auc_score
 
 
 class TestSklearn(unittest.TestCase):
@@ -68,3 +68,42 @@ def test_weights(self):
         np.testing.assert_array_almost_equal(y_swd, y, decimal=6)
         self.assertGreater(np.sum(np.abs(y_cw - y)), 0.001)
         np.testing.assert_array_almost_equal(y_cw, y_sw, decimal=6)
+
+    def test_classifier(self):
+        params = [{'num_threads': 1, 'time_limit': 0.0,'iter_limit': 1000, 'random_state': 42},
+                  {'num_threads': 2, 'time_limit': 0.0,'iter_limit': 1000, 'random_state': 42},]
+        classifiers = [NonlinearLogisticRegressor, SymbolicClassifier, FuzzyRegressor, FuzzyClassifier]
+        for p in params:
+            for clf in classifiers:
+                model = clf(**p)
+                model.fit(self.X_train, self.y_train)
+                y = model.predict(self.X_test)
+                yp = model.predict_proba(self.X_test)
+                np.testing.assert_equal(y.shape, self.y_test.shape)
+                np.testing.assert_equal(yp.shape, (self.y_test.shape[0], 2))
+
+    def test_classifier_cv(self):
+        cv = {'n':2, 'cv_params':{}, 'select':'mean', 'opt_params':{'method': 'L-BFGS-B'}, 'opt_metric':make_scorer(log_loss, greater_is_better=False, needs_proba=True)}
+        params = [{'num_threads': 1, 'time_limit': 0.0,'iter_limit': 1000, 'random_state': 42, 'cv_params' : cv},
+                  {'num_threads': 2, 'time_limit': 0.0,'iter_limit': 1000, 'random_state': 42, 'cv_params' : cv},]
+        classifiers = [NonlinearLogisticRegressor, SymbolicClassifier, FuzzyRegressor, FuzzyClassifier]
+        for p in params:
+            for clf in classifiers:
+                model = clf(**p)
+                model.fit(self.X_train, self.y_train)
+                y = model.predict(self.X_test)
+                yp = model.predict_proba(self.X_test)
+                np.testing.assert_equal(y.shape, self.y_test.shape)
+                np.testing.assert_equal(yp.shape, (self.y_test.shape[0], 2))
+                if clf in [SymbolicClassifier, FuzzyClassifier]:
+                    self.assertEqual(len(model.estimators_), 1)
+                    model = model.estimators_[0]
+                equations = model.get_models()[:2]
+                for eq in equations:
+                    self.assertTrue(hasattr(eq, 'cv_score'))
+                    eq.fit(self.X_train, self.y_train)
+                    y = eq.predict(self.X_test)
+                    yp = eq.predict_proba(self.X_test)
+                    np.testing.assert_equal(y.shape, self.y_test.shape)
+                    np.testing.assert_equal(yp.shape, (self.y_test.shape[0], 2))
+