为什么我在所有 10 次折叠中得到相同的测试准确度和平衡准确度值？

Question

For each fold test accuracy and balanced test accuracy is different, but the values are same. For example, Fold 1, test accuracy is 86, balanced test accuracy is 86. For Fold 2, test accuracy is 90, balanced test accuracy is 90 For fold 3, test accuracy is 70.555, test accuracy is 70.555...Here is my code

fold_no = 1
reports = []
accuracies = []
sensitivities = []
specificities = []
test_accuracy = []
for train, test in kfold.split(X_train, y_train):

  model = Sequential()
  model.add(Conv3D(128, kernel_size=(3, 3, 3))
  model.add(Flatten())
  model.add(Dense(256, activation='relu', kernel_regularizer='l2'))
  model.add(Dense(4096, activation='relu', kernel_regularizer='l2')) 
  model.add(Dropout(0.3))
  model.add(Dense(1, activation='sigmoid', kernel_regularizer='l2'))

  # Compile the model
  model.compile(loss=tensorflow.keras.losses.mean_squared_error,
                optimizer=tensorflow.keras.optimizers.Adam(learning_rate=learning_rate),
                metrics=['accuracy'])```

  history = model.fit(X_train[train], y_train[train],
                      batch_size=batch_size,
                      epochs=no_epochs,
                      verbose=verbosity, validation_data=(X_train[test], y_train[test]))

  # Compute the classification report for the testing set
  y_pred = model.predict(X_test, verbose = 0)
  c = model.evaluate(X_test, y_test)
  test_accuracy.append(c[1])
  report = classification_report(y_test, (y_pred>0.5), output_dict=True)
  from sklearn.metrics import balanced_accuracy_score
  bal_acc=balanced_accuracy_score(y_test,(y_pred>0.5))
  print("balenced acc is " + str(bal_acc))

  # Extract the sensitivity and specificity values from the report
  sensitivity = report["1"]["recall"]
  specificity = report["0"]["recall"]
  sensitivities.append(sensitivity)
  specificities.append(specificity)

  print(specificity))  
  print(sensitivity))

Answer 1

When classes are balanced to begin with, balanced accuracy and accuracy are the same:

from sklearn.metrics import accuracy_score, balanced_accuracy_score

y_true = [0, 0, 0, 0, 1, 1, 1, 1]    # 4 negatives, 4 positives
y_pred = [0, 0, 1, 0, 1, 0, 1, 1]

print(accuracy_score(y_true, y_pred), balanced_accuracy_score(y_true, y_pred))
# 0.75 0.75