如何在给定代码场景下计算ANN的测试集准确率？

Hey there! Let's get your ANN's test accuracy sorted out. I notice a couple of key adjustments needed in your code, plus we'll fill in the missing pieces to calculate accuracy properly.

Key Issues to Address #

• You're using KerasRegressor for a classification task: Accuracy is a metric for classification problems (predicting categories), but KerasRegressor is designed for regression (predicting continuous values). For classification, you need KerasClassifier instead.
• Missing data split: To evaluate test set accuracy, you need to separate your dataset into training and unseen test subsets.
• Incomplete model setup: Your model needs an output layer and loss function tailored to classification (not regression).

Corrected Code Example #

Here's a revised version of your code that calculates test set accuracy correctly. I'll assume you're working on a binary classification task (adjust for multi-class if needed):

import numpy as np
import pandas as pd
from keras.models import Sequential
from keras.layers import Dense
from keras.wrappers.scikit_learn import KerasClassifier  # Switched from KerasRegressor
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline
from sklearn.metrics import accuracy_score

# Load your dataset (replace with your actual data path/loading logic)
data = pd.read_csv("your_dataset.csv")
X = data.drop("target_column", axis=1).values  # Features
y = data["target_column"].values               # Target labels

# Split data into training (80%) and test (20%) sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Define your classification ANN model
def create_classifier():
    model = Sequential()
    model.add(Dense(12, input_dim=X_train.shape[1], activation='relu'))  # Input + hidden layer
    model.add(Dense(8, activation='relu'))                               # Second hidden layer
    # Binary classification output layer (adjust for multi-class if needed)
    model.add(Dense(1, activation='sigmoid'))
    # Compile with classification-specific loss and metrics
    model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
    return model

# Create a pipeline with scaling and the classifier
estimator = Pipeline([
    ('scaler', StandardScaler()),  # Normalize features
    ('classifier', KerasClassifier(build_fn=create_classifier, epochs=100, batch_size=5, verbose=0))
])

# Train the model on training data
estimator.fit(X_train, y_train)

# Predict labels for the test set
y_pred = estimator.predict(X_test)

# Calculate and print test set accuracy
test_accuracy = accuracy_score(y_test, y_pred)
print(f"Test Set Accuracy: {test_accuracy:.4f}")

Additional Notes #

• Multi-class adjustment: If you're working with more than 2 classes, modify the output layer to use Dense(n_classes, activation='softmax') and change the loss function to categorical_crossentropy (ensure your target labels are one-hot encoded, or use sparse_categorical_crossentropy for integer labels).
• Regression check: If you're actually doing regression (predicting continuous values), accuracy isn't the right metric—use metrics like Mean Absolute Error (MAE) or R-squared instead.
• Python 3.5 compatibility: Since you're using Python 3.5, make sure you install compatible library versions: Keras ≤ 2.3.1, scikit-learn ≤ 0.20.3, and TensorFlow ≤ 1.15 (newer versions dropped support for Python 3.5).

内容的提问来源于stack exchange，提问作者Kashish Goyal