Skip to content

Decision tree

Decision Tree classification implementation for SciREX.

This module provides a Decision Tree implementation using scikit-learn with automatic parameter tuning using grid search. The implementation focuses on both accuracy and interpretability.

Mathematical Background

Decision Trees recursively partition the feature space using:

  1. Splitting Criteria:
  2. Gini Impurity: 1 - ∑ᵢpᵢ²

  3. Entropy: -∑ᵢpᵢlog(pᵢ) where pᵢ is the proportion of class i in the node

  4. Information Gain: IG(parent, children) = I(parent) - ∑(nⱼ/n)I(childⱼ) where I is impurity measure (Gini or Entropy)

  5. Tree Pruning: Cost-Complexity: Rα(T) = R(T) + α|T| where R(T) is tree error, |T| is tree size, α is complexity parameter

Key Features
  • Automatic parameter optimization
  • Multiple splitting criteria
  • Built-in tree visualization
  • Pruning capabilities
  • Feature importance estimation
References

[1] Breiman, L., et al. (1984). Classification and Regression Trees [2] Quinlan, J. R. (1986). Induction of Decision Trees [3] Hastie, T., et al. (2009). Elements of Statistical Learning, Ch. 9

DecisionTreeClassifier

Bases: Classification

Decision Tree with automatic parameter tuning.

This implementation includes automatic selection of optimal parameters using grid search with cross-validation. It balances model complexity with performance through pruning and parameter optimization.

Attributes:

Name Type Description
cv

Number of cross-validation folds
best_params Optional[Dict[str, Any]]

Best parameters found by grid search
model Optional[Any]

Fitted DecisionTreeClassifier instance
Example

classifier = DecisionTreeClassifier(cv=5) X_train = np.array([[1, 2], [2, 3], [3, 4]]) y_train = np.array([0, 0, 1]) classifier.fit(X_train, y_train) print(classifier.best_params)

Source code in scirex/core/ml/supervised/classification/decision_tree.py 
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
class DecisionTreeClassifier(Classification):
    """Decision Tree with automatic parameter tuning.

    This implementation includes automatic selection of optimal parameters
    using grid search with cross-validation. It balances model complexity
    with performance through pruning and parameter optimization.

    Attributes:
        cv: Number of cross-validation folds
        best_params: Best parameters found by grid search
        model: Fitted DecisionTreeClassifier instance

    Example:
        >>> classifier = DecisionTreeClassifier(cv=5)
        >>> X_train = np.array([[1, 2], [2, 3], [3, 4]])
        >>> y_train = np.array([0, 0, 1])
        >>> classifier.fit(X_train, y_train)
        >>> print(classifier.best_params)
    """

    def __init__(self, cv: int = 5, **kwargs: Any) -> None:
        """Initialize Decision Tree classifier.

        Args:
            cv: Number of cross-validation folds. Defaults to 5.
            **kwargs: Additional keyword arguments passed to parent class.

        Notes:
            The classifier uses GridSearchCV for parameter optimization,
            searching over different tree depths, splitting criteria,
            and minimum sample thresholds.
        """
        super().__init__("decision_tree", **kwargs)
        self.cv = cv
        self.best_params: Optional[Dict[str, Any]] = None

    def fit(self, X: np.ndarray, y: np.ndarray) -> None:
        """Fit Decision Tree model with parameter tuning.

        Performs grid search over tree parameters to find optimal
        model configuration using cross-validation.

        Args:
            X: Training feature matrix of shape (n_samples, n_features)
            y: Training labels of shape (n_samples,)

        Notes:
            The grid search optimizes over:
            - Splitting criterion (gini vs entropy)
            - Maximum tree depth
            - Minimum samples for splitting
            - Minimum samples per leaf
            - Maximum features considered per split
        """
        param_grid = {
            "criterion": ["gini", "entropy"],
            "max_depth": [3, 5, 7, 9, None],
            "min_samples_split": [2, 5, 10],
            "min_samples_leaf": [1, 2, 4],
            "max_features": ["sqrt", "log2", None],
        }

        base_model = DTC(random_state=self.random_state)

        grid_search = GridSearchCV(
            base_model, param_grid, cv=self.cv, scoring="accuracy", n_jobs=-1
        )

        grid_search.fit(X, y)

        self.best_params = grid_search.best_params_
        self.model = grid_search.best_estimator_

        print(f"Best parameters found: {self.best_params}")
        print(f"Best cross-validation score: {grid_search.best_score_:.4f}")

    def get_model_params(self) -> Dict[str, Any]:
        """Get parameters of the fitted model.

        Returns:
            Dictionary containing:
                - model_type: Type of classifier
                - best_params: Best parameters found by grid search
                - cv: Number of cross-validation folds used
        """
        return {
            "model_type": self.model_type,
            "best_params": self.best_params,
            "cv": self.cv,
        }

    def get_feature_importance(self) -> Dict[str, float]:
        """Get feature importance scores.

        Returns:
            Dictionary mapping feature indices to importance scores

        Raises:
            ValueError: If model hasn't been fitted yet

        Notes:
            Feature importance is computed based on the decrease in
            impurity (Gini or entropy) brought by each feature across
            all tree splits.
        """
        if self.model is None:
            raise ValueError("Model must be fitted before getting feature importance")

        importance_dict = {}
        for idx, importance in enumerate(self.model.feature_importances_):
            importance_dict[f"feature_{idx}"] = importance

        return importance_dict

    # Add these methods to decision_tree.py

    def predict(self, X: np.ndarray) -> np.ndarray:
        """Predict class labels for samples in X.

        Args:
            X: Test samples of shape (n_samples, n_features)

        Returns:
            Array of predicted class labels

        Raises:
            ValueError: If model hasn't been fitted yet
        """
        if self.model is None:
            raise ValueError("Model must be fitted before prediction")
        return self.model.predict(X)

    def predict_proba(self, X: np.ndarray) -> np.ndarray:
        """Predict class probabilities for samples in X.

        Args:
            X: Test samples of shape (n_samples, n_features)

        Returns:
            Array of shape (n_samples, n_classes) with class probabilities

        Raises:
            ValueError: If model hasn't been fitted yet
        """
        if self.model is None:
            raise ValueError("Model must be fitted before prediction")
        return self.model.predict_proba(X)

    def evaluate(self, X_test: np.ndarray, y_test: np.ndarray) -> Dict[str, float]:
        """Evaluate model performance on test data.

        Args:
            X_test: Test features of shape (n_samples, n_features)
            y_test: True labels of shape (n_samples,)

        Returns:
            Dictionary containing evaluation metrics:
                - accuracy: Overall classification accuracy
                - precision: Precision score (micro-averaged)
                - recall: Recall score (micro-averaged)
                - f1_score: F1 score (micro-averaged)
        """
        from sklearn.metrics import (
            accuracy_score,
            precision_score,
            recall_score,
            f1_score,
        )

        if self.model is None:
            raise ValueError("Model must be fitted before evaluation")

        y_pred = self.predict(X_test)

        return {
            "accuracy": accuracy_score(y_test, y_pred),
            "precision": precision_score(y_test, y_pred, average="weighted"),
            "recall": recall_score(y_test, y_pred, average="weighted"),
            "f1_score": f1_score(y_test, y_pred, average="weighted"),
        }

__init__(cv=5, **kwargs)

Initialize Decision Tree classifier.

Parameters:

Name Type Description Default
cv int

Number of cross-validation folds. Defaults to 5.

 5
**kwargs Any

Additional keyword arguments passed to parent class.

 {}
Notes

The classifier uses GridSearchCV for parameter optimization, searching over different tree depths, splitting criteria, and minimum sample thresholds.

Source code in scirex/core/ml/supervised/classification/decision_tree.py 
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
def __init__(self, cv: int = 5, **kwargs: Any) -> None:
    """Initialize Decision Tree classifier.

    Args:
        cv: Number of cross-validation folds. Defaults to 5.
        **kwargs: Additional keyword arguments passed to parent class.

    Notes:
        The classifier uses GridSearchCV for parameter optimization,
        searching over different tree depths, splitting criteria,
        and minimum sample thresholds.
    """
    super().__init__("decision_tree", **kwargs)
    self.cv = cv
    self.best_params: Optional[Dict[str, Any]] = None

evaluate(X_test, y_test)

Evaluate model performance on test data.

Parameters:

Name Type Description Default
X_test ndarray

Test features of shape (n_samples, n_features)

 required
y_test ndarray

True labels of shape (n_samples,)

 required

Returns:

Type Description
Dict[str, float]

Dictionary containing evaluation metrics: - accuracy: Overall classification accuracy - precision: Precision score (micro-averaged) - recall: Recall score (micro-averaged) - f1_score: F1 score (micro-averaged)
Source code in scirex/core/ml/supervised/classification/decision_tree.py 
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
def evaluate(self, X_test: np.ndarray, y_test: np.ndarray) -> Dict[str, float]:
    """Evaluate model performance on test data.

    Args:
        X_test: Test features of shape (n_samples, n_features)
        y_test: True labels of shape (n_samples,)

    Returns:
        Dictionary containing evaluation metrics:
            - accuracy: Overall classification accuracy
            - precision: Precision score (micro-averaged)
            - recall: Recall score (micro-averaged)
            - f1_score: F1 score (micro-averaged)
    """
    from sklearn.metrics import (
        accuracy_score,
        precision_score,
        recall_score,
        f1_score,
    )

    if self.model is None:
        raise ValueError("Model must be fitted before evaluation")

    y_pred = self.predict(X_test)

    return {
        "accuracy": accuracy_score(y_test, y_pred),
        "precision": precision_score(y_test, y_pred, average="weighted"),
        "recall": recall_score(y_test, y_pred, average="weighted"),
        "f1_score": f1_score(y_test, y_pred, average="weighted"),
    }

fit(X, y)

Fit Decision Tree model with parameter tuning.

Performs grid search over tree parameters to find optimal model configuration using cross-validation.

Parameters:

Name Type Description Default
X ndarray

Training feature matrix of shape (n_samples, n_features)

 required
y ndarray

Training labels of shape (n_samples,)

 required
Notes

The grid search optimizes over: - Splitting criterion (gini vs entropy) - Maximum tree depth - Minimum samples for splitting - Minimum samples per leaf - Maximum features considered per split

Source code in scirex/core/ml/supervised/classification/decision_tree.py 
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
def fit(self, X: np.ndarray, y: np.ndarray) -> None:
    """Fit Decision Tree model with parameter tuning.

    Performs grid search over tree parameters to find optimal
    model configuration using cross-validation.

    Args:
        X: Training feature matrix of shape (n_samples, n_features)
        y: Training labels of shape (n_samples,)

    Notes:
        The grid search optimizes over:
        - Splitting criterion (gini vs entropy)
        - Maximum tree depth
        - Minimum samples for splitting
        - Minimum samples per leaf
        - Maximum features considered per split
    """
    param_grid = {
        "criterion": ["gini", "entropy"],
        "max_depth": [3, 5, 7, 9, None],
        "min_samples_split": [2, 5, 10],
        "min_samples_leaf": [1, 2, 4],
        "max_features": ["sqrt", "log2", None],
    }

    base_model = DTC(random_state=self.random_state)

    grid_search = GridSearchCV(
        base_model, param_grid, cv=self.cv, scoring="accuracy", n_jobs=-1
    )

    grid_search.fit(X, y)

    self.best_params = grid_search.best_params_
    self.model = grid_search.best_estimator_

    print(f"Best parameters found: {self.best_params}")
    print(f"Best cross-validation score: {grid_search.best_score_:.4f}")

get_feature_importance()

Get feature importance scores.

Returns:

Type Description
Dict[str, float]

Dictionary mapping feature indices to importance scores

Raises:

Type Description
ValueError

If model hasn't been fitted yet
Notes

Feature importance is computed based on the decrease in impurity (Gini or entropy) brought by each feature across all tree splits.

Source code in scirex/core/ml/supervised/classification/decision_tree.py 
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
def get_feature_importance(self) -> Dict[str, float]:
    """Get feature importance scores.

    Returns:
        Dictionary mapping feature indices to importance scores

    Raises:
        ValueError: If model hasn't been fitted yet

    Notes:
        Feature importance is computed based on the decrease in
        impurity (Gini or entropy) brought by each feature across
        all tree splits.
    """
    if self.model is None:
        raise ValueError("Model must be fitted before getting feature importance")

    importance_dict = {}
    for idx, importance in enumerate(self.model.feature_importances_):
        importance_dict[f"feature_{idx}"] = importance

    return importance_dict

get_model_params()

Get parameters of the fitted model.

Returns:

Type Description
Dict[str, Any]

Dictionary containing: - model_type: Type of classifier - best_params: Best parameters found by grid search - cv: Number of cross-validation folds used
Source code in scirex/core/ml/supervised/classification/decision_tree.py 
144
145
146
147
148
149
150
151
152
153
154
155
156
157
def get_model_params(self) -> Dict[str, Any]:
    """Get parameters of the fitted model.

    Returns:
        Dictionary containing:
            - model_type: Type of classifier
            - best_params: Best parameters found by grid search
            - cv: Number of cross-validation folds used
    """
    return {
        "model_type": self.model_type,
        "best_params": self.best_params,
        "cv": self.cv,
    }

predict(X)

Predict class labels for samples in X.

Parameters:

Name Type Description Default
X ndarray

Test samples of shape (n_samples, n_features)

 required

Returns:

Type Description
ndarray

Array of predicted class labels

Raises:

Type Description
ValueError

If model hasn't been fitted yet
Source code in scirex/core/ml/supervised/classification/decision_tree.py 
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
def predict(self, X: np.ndarray) -> np.ndarray:
    """Predict class labels for samples in X.

    Args:
        X: Test samples of shape (n_samples, n_features)

    Returns:
        Array of predicted class labels

    Raises:
        ValueError: If model hasn't been fitted yet
    """
    if self.model is None:
        raise ValueError("Model must be fitted before prediction")
    return self.model.predict(X)

predict_proba(X)

Predict class probabilities for samples in X.

Parameters:

Name Type Description Default
X ndarray

Test samples of shape (n_samples, n_features)

 required

Returns:

Type Description
ndarray

Array of shape (n_samples, n_classes) with class probabilities

Raises:

Type Description
ValueError

If model hasn't been fitted yet
Source code in scirex/core/ml/supervised/classification/decision_tree.py 
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
def predict_proba(self, X: np.ndarray) -> np.ndarray:
    """Predict class probabilities for samples in X.

    Args:
        X: Test samples of shape (n_samples, n_features)

    Returns:
        Array of shape (n_samples, n_classes) with class probabilities

    Raises:
        ValueError: If model hasn't been fitted yet
    """
    if self.model is None:
        raise ValueError("Model must be fitted before prediction")
    return self.model.predict_proba(X)