update Part_A.py

Part_A.py

@@ -22,6 +22,7 @@ def cross_validation(classifier, features, labels, scoring_metrics):
     return scores.mean()
 
 def main():
+    best_classifier_name = ""
     best_classifier_average = 0
 
     # Read the training data
@@ -38,6 +39,8 @@ def main():
     scaler = StandardScaler()
     train_features_scaled = scaler.fit_transform(train_features)
 
+    # The following commented out code was used to find optimal hyperparameters
+    '''
     # Define the hyperparameters for the Bayes search
     parameters = {
         "solver" : ["lbfgs", "sgd", "adam"],
@@ -49,36 +52,42 @@ def main():
     search = BayesSearchCV(MLPClassifier(max_iter = 10000), parameters, n_iter = 50, n_jobs = -1, cv = 5, scoring = "accuracy").fit(train_features_scaled, train_labels)
     print(f"Best Score: {search.best_score_}")
     print(f"Best Hyperparameters: {search.best_params_}")
-
+    '''
 
     # Train the classifiers
     classifiers = [
-        LogisticRegression(max_iter = 10000, solver = "newton-cg", C = 9.416).fit(train_features_scaled, train_labels),
-        MLPClassifier(max_iter = 10000, solver = "adam", activation = "tanh", learning_rate = "constant").fit(train_features_scaled, train_labels),
-        SVC(C = 7.989999999999979, kernel = "linear").fit(train_features_scaled, train_labels),
-        #RandomForestClassifier(n_estimators = 418, max_depth = 5).fit(train_features_scaled, train_labels),
-        #DecisionTreeClassifier(max_features = "sqrt", criterion = "gini", max_depth = 19).fit(train_features_scaled, train_labels),
-        #KNeighborsClassifier(n_neighbors = 4, n_jobs = -1, leaf_size = 68, metric = "manhattan", weights = "distance", algorithm = "kd_tree").fit(train_features_scaled, train_labels)
+        ("Logistical Regression" , LogisticRegression(max_iter = 10000, solver = "newton-cg", C = 9.416).fit(train_features_scaled, train_labels)),
+        ("Multi-layer Perceptron" , MLPClassifier(max_iter = 10000, solver = "adam", activation = "tanh", learning_rate = "constant").fit(train_features_scaled, train_labels)),
+        ("C-Support Vector" , SVC(C = 7.989999999999979, kernel = "linear").fit(train_features_scaled, train_labels)),
+
+        # These algorithms were optimised and trained but were found to be overfitting
+        #("Random Forest" , RandomForestClassifier(n_estimators = 418, max_depth = 5).fit(train_features_scaled, train_labels)),
+        #("Decision Tree" , DecisionTreeClassifier(max_features = "sqrt", criterion = "gini", max_depth = 19).fit(train_features_scaled, train_labels)),
+        #("K-Nearest Neighbours" , KNeighborsClassifier(n_neighbors = 4, n_jobs = -1, leaf_size = 68, metric = "manhattan", weights = "distance", algorithm = "kd_tree").fit(train_features_scaled, train_labels))
     ]
 
     # Evaluate the performance of the trained classifiers
-    for classifier in classifiers:
+    for classifier_name, classifier in classifiers:
+        print(f"Training {classifier_name}...")
         train_predicted_labels = classifier.predict(train_features_scaled)
         train_accuracy = accuracy_score(train_labels, train_predicted_labels)
         print(f"Training Accuracy:", train_accuracy)
         cv_accuracy = cross_validation(classifier, train_features_scaled, train_labels, "accuracy")
         print(f"Cross-Validated Accuracy: {cv_accuracy}")
-        #cv_f1 = cross_validation(classifier, train_features_scaled, train_labels, "f1")
+        cv_f1 = cross_validation(classifier, train_features_scaled, train_labels, "f1")
         cv_f1 = cv_accuracy
-        #print(f"Cross-Validated F1-Score: {cv_f1}")
+        print(f"Cross-Validated F1-Score: {cv_f1}")
         cv_average = ((cv_accuracy + cv_f1) / 2)
-        print(f"Average Score: {cv_average}")
+        print(f"Average Score: {cv_average} \n")
 
         # Update the best classifier if the current classifier has a better average score
         if (best_classifier_average < cv_average):
+            best_classifier_name = classifier_name
             best_classifier = classifier
             best_classifier_average = cv_average
 
+    print(f"The best classifier is {best_classifier_name}.\n")
+
     # Read the training data
     test_features = pd.read_csv("TestingDataBinary.csv", header = None)
 
@@ -93,5 +102,7 @@ def main():
     test_results["Predicted Labels"] = test_predicted_labels
     test_results.to_csv("TestingResultsBinary.csv", header = False, index = False)
 
+    print("Test data was trained and new labels have been predicted.")
+
 if __name__ == "__main__":
     main()
\ No newline at end of file