Hey there! Let's figure out why you're getting that "Unknown label type: 'continuous-multioutput'" error and fix your code so it runs smoothly. I'll break down the issues step by step and show you the corrected version.

Key Issues Causing the Error #

1. Mismatched Parameter Order in train_predict Call
   You defined train_predict to take parameters in this order: learner, sample_size, X_train, X_test, y_train, y_test, but when you call it in your loop, you're passing: train_predict(clf, samples, X_train, y_train, X_test, y_test). This swaps X_test and y_train, meaning you're trying to train the model using y_train as features and X_test as labels—total chaos! This is almost certainly the main reason for the weird label type error.

2. Wrong Order in Metric Calculations
   Both accuracy_score and fbeta_score require true labels first, then predictions—you're passing features instead of predictions, which makes no sense for classification metrics.

3. Spelling Typo
   You have resutts['f_test'] (missing an 'l') instead of results['f_test']—this would throw a NameError once you fix the other issues.

4. Missing Import
   You're using time() but haven't imported the time module.

5. Potential Label Format Issue
   Even if you fix the parameter order, if your y_train/y_test are 2D arrays (e.g., shape (n_samples, 1)) or continuous values instead of discrete 0/1 labels, you'll still get the label type error. Classification models need 1D discrete class labels.

Corrected Full Code #

First, make sure your labels are properly formatted (add this before your main code):

# Example: Convert salary labels to binary 0/1 (adjust based on your dataset)
# Assuming your training data has a 'salary' column with values like '>50K'/'<=50K'
y_train = (train_data['salary'] == '>50K').astype(int)
y_test = (test_data['salary'] == '>50K').astype(int)

# Ensure labels are 1D arrays (not 2D)
y_train = y_train.values.ravel()
y_test = y_test.values.ravel()

Now the corrected main code:

import time
from sklearn.metrics import fbeta_score, accuracy_score
from sklearn.tree import DecisionTreeClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.svm import SVC

def train_predict(learner, sample_size, X_train, X_test, y_train, y_test):
    results = {}
    
    # Train the learner
    start = time.time()
    learner.fit(X_train[:sample_size], y_train[:sample_size])  # Use matching sample size for y_train
    end = time.time()
    results['train_time'] = end - start
    
    # Make predictions
    start = time.time()
    predictions_test = learner.predict(X_test)
    predictions_train = learner.predict(X_train[:sample_size])
    end = time.time()
    results['pred_time'] = end - start
    
    # Calculate metrics (correct order: true labels first, predictions second)
    results['acc_train'] = accuracy_score(y_train[:sample_size], predictions_train)
    results['acc_test'] = accuracy_score(y_test, predictions_test)
    results['f_train'] = fbeta_score(y_train[:sample_size], predictions_train, beta=1)
    results['f_test'] = fbeta_score(y_test, predictions_test, beta=1)  # Fixed typo here
    
    print(f"{learner.__class__.__name__} trained on {sample_size} samples.")
    return results

# Initialize classifiers (add random_state for reproducibility)
clf_A = DecisionTreeClassifier(random_state=42)
clf_B = GaussianNB()
clf_C = SVC(random_state=42)

# Define sample sizes
samples_100 = len(X_train)
samples_10 = int(len(X_train) * 0.1)
samples_1 = int(len(X_train) * 0.01)

# Collect results
results = {}
for clf in [clf_A, clf_B, clf_C]:
    clf_name = clf.__class__.__name__
    results[clf_name] = {}
    for i, samples in enumerate([samples_1, samples_10, samples_100]):
        # Fix parameter order here to match the function definition
        results[clf_name][i] = train_predict(clf, samples, X_train, X_test, y_train, y_test)

# If you don't have the custom vs.evaluate function, print results manually:
for clf_name, clf_results in results.items():
    print(f"\n--- {clf_name} ---")
    for sample_idx, metrics in clf_results.items():
        sample_size = [samples_1, samples_10, samples_100][sample_idx]
        print(f"Sample size: {sample_size}")
        print(f"Train Accuracy: {metrics['acc_train']:.4f}, Train F1: {metrics['f_train']:.4f}")
        print(f"Test Accuracy: {metrics['acc_test']:.4f}, Test F1: {metrics['f_test']:.4f}")
        print(f"Train Time: {metrics['train_time']:.4f}s, Predict Time: {metrics['pred_time']:.4f}s")

Additional Tips for Beginners #

• Always check parameter orders: Sklearn functions are strict about argument order—double-check the docs if you're unsure.
• Reproducibility: Add random_state to classifiers so you get the same results every time you run the code.
• SVC can be slow: For large datasets, SVC with default settings might take a long time. You could try SVC(kernel='linear') for faster training, or use a different classifier like Logistic Regression if speed is an issue.
• Check data shapes: Use print(X_train.shape, y_train.shape) to make sure your features and labels have compatible shapes (X should be (n_samples, n_features), y should be (n_samples,)).

内容的提问来源于stack exchange，提问作者user11534866