Add nodes, leaves and depth average

odte/Odte.py

@@ -83,8 +83,21 @@ class Odte(BaseEnsemble, ClassifierMixin):  # type: ignore
         self.subspaces_: List[Tuple[int, ...]] = []
         result = self._train(X, y, sample_weight)
         self.estimators_, self.subspaces_ = tuple(zip(*result))  # type: ignore
+        self._compute_metrics()
         return self
 
+    def _compute_metrics(self) -> None:
+        tdepth = tnodes = tleaves = 0
+        for estimator in self.estimators_:
+            nodes, leaves = estimator.nodes_leaves()
+            depth = estimator.depth_
+            tdepth += depth
+            tnodes += nodes
+            tleaves += leaves
+        self.depth_ = tdepth / self.n_estimators
+        self.leaves_ = tleaves / self.n_estimators
+        self.nodes_ = tnodes / self.n_estimators
+
     @staticmethod
     def _parallel_build_tree(
         base_estimator_: Stree,
@@ -228,3 +241,7 @@ class Odte(BaseEnsemble, ClassifierMixin):  # type: ignore
             for i in range(n_samples):
                 result[i, predictions[i]] += 1
         return result / self.n_estimators
+
+    def nodes_leaves(self) -> list(float, float):
+        check_is_fitted(self, "estimators_")
+        return self.nodes_, self.leaves_

odte/tests/Odte_tests.py

@@ -3,7 +3,7 @@ import unittest
 import os
 import random
 import warnings
-from sklearn.exceptions import ConvergenceWarning
+from sklearn.exceptions import ConvergenceWarning, NotFittedError
 
 from odte import Odte
 from stree import Stree
@@ -191,3 +191,27 @@ class Odte_test(unittest.TestCase):
         from sklearn.utils.estimator_checks import check_estimator
 
         check_estimator(Odte())
+
+    def test_nodes_leaves_not_fitted(self):
+        tclf = Odte(
+            base_estimator=Stree(),
+            random_state=self._random_state,
+            n_estimators=3,
+        )
+        with self.assertRaises(NotFittedError):
+            tclf.nodes_leaves()
+
+    def test_nodes_leaves_depth(self):
+        tclf = Odte(
+            base_estimator=Stree(),
+            random_state=self._random_state,
+            n_estimators=3,
+        )
+        X, y = load_dataset(self._random_state, n_features=16, n_samples=500)
+        tclf.fit(X, y)
+        self.assertAlmostEqual(6.0, tclf.depth_)
+        self.assertAlmostEqual(9.333333333333334, tclf.leaves_)
+        self.assertAlmostEqual(17.666666666666668, tclf.nodes_)
+        nodes, leaves = tclf.nodes_leaves()
+        self.assertAlmostEqual(9.333333333333334, leaves)
+        self.assertAlmostEqual(17.666666666666668, nodes)