Update examples for MhaSelector and custom objective function

Author: anh9895

examples/wrapper/exam_mha.py

@@ -6,15 +6,15 @@
 
 from mafese import get_dataset, MhaSelector
 
-data = get_dataset("aniso")
+data = get_dataset("balance")
 data.split_train_test(test_size=0.2)
 
 
 selector = MhaSelector(problem="classification", obj_name="F1S",
                        estimator="svm", estimator_paras=None,
                        optimizer="BaseGA", optimizer_paras={"epoch": 100, "pop_size": 20, "name": "GA"},
                        mode='single', n_workers=None, termination=None,
-                       seed=None, verbose=True)
+                       seed=42, verbose=True)
 selector.fit(data.X_train, data.y_train)
 
 # Transform test data

examples/wrapper/exam_mha_custom_obj.py

@@ -0,0 +1,67 @@
+#!/usr/bin/env python
+# Created by "Thieu" at 15:35, 01/06/2025 ----------%
+#       Email: [email protected]            %                                                    
+#       Github: https://github.com/thieu1995        %                         
+# --------------------------------------------------%
+
+import numpy as np
+from mafese import get_dataset, MhaSelector, FeatureSelectionProblem
+
+
+# Define a custom callback objective function
+def my_custom_problem(**kwargs):
+
+    class MyProblem(FeatureSelectionProblem):
+        # Please check out the FeatureSelectionProblem class in mafese/utils/mealpy_util.py to know which attributes are existing in the object
+
+        def obj_func(self, solution):
+            # Decode the solution to get the selected features
+            x = self.decode_solution(solution)["my_var"]  # Please don't change this line, it is used to decode the solution
+            cols = np.flatnonzero(x)  # Get columns where features are selected
+
+            # In self.data object we have train and test data, we can just use them directly
+
+            # Fit the estimator on the selected features
+            self.estimator.fit(self.data.X_train[:, cols], self.data.y_train)
+
+            # Predict on the test set
+            y_valid_pred = self.estimator.predict(self.data.X_test[:, cols])
+
+            # Evaluate using a custom metric (e.g., F1 score)
+            evaluator = self.metric_class(self.data.y_test, y_valid_pred)
+            obj = evaluator.get_metric_by_name("NPV", paras=self.obj_paras)["NPV"]  # Change "NPV" to your desired metric name
+
+            # Calculate fitness
+            fitness = self.fit_weights[0] * obj + self.fit_weights[1] * self.fit_sign * (np.sum(x) / self.n_dims)
+            return [fitness, obj, int(np.sum(x))]  # Return fitness, objective value, and number of selected features
+
+    return MyProblem(**kwargs)
+
+
+data = get_dataset("aniso")
+data.split_train_test(test_size=0.2)
+
+selector = MhaSelector(problem="classification", obj_name="F1S",
+                       estimator="svm", estimator_paras=None,
+                       optimizer="BaseGA", optimizer_paras={"epoch": 100, "pop_size": 20, "name": "GA"},
+                       mode='single', n_workers=None, termination=None,
+                       seed=42, verbose=True)
+selector.fit(data.X_train, data.y_train, fs_problem=my_custom_problem)
+
+# Transform test data
+X_selected = selector.transform(data.X_test)
+print(f"Original Dataset: {data.X_train.shape}")
+print(f"Selected dataset: {X_selected.shape}")
+
+# Get some information
+print(selector.get_best_information())
+print(selector.selected_feature_masks)
+print(selector.selected_feature_solution)
+print(selector.selected_feature_indexes)
+
+# Predict with new selected features
+res1 = selector.evaluate(estimator=None, estimator_paras=None, data=data, metrics=["AS", "PS", "RS"])
+# AS: Accuracy score, PS: precision score, RS: recall score
+print(res1)
+
+