SVM测试集评分问题：使用scikit-learn时遭遇ValueError

Hey there, I spotted the issue in your code that's triggering that ValueError—let's get it sorted out!

The Root Cause #

You made a small but critical mistake when defining your test set labels:

y_test_1 = dataset[:,15:16]  # Wrong! You're using the training dataset here

Instead of pulling labels from your test dataset (test_dataset_1), you're accidentally using the training dataset (dataset). This means your test features (X_test_1) and test labels (y_test_1) almost certainly have different numbers of samples, which scikit-learn can't handle when calculating the model score.

Corrected Code #

Here's the fixed version of your full code:

from sklearn.svm import SVC
import numpy

dataset = numpy.loadtxt("training.txt", delimiter="\t")
X = dataset[:,0:15]
y = dataset[:,15:16]
y = y.ravel()

test_dataset_1 = numpy.loadtxt("test_14-15.txt", delimiter="\t")
X_test_1 = test_dataset_1[:,0:15]
y_test_1 = test_dataset_1[:,15:16]  # Fixed: use test_dataset_1 here
y_test_1 = y_test_1.ravel()

model = SVC(kernel='linear', C=75)
model.fit(X, y)
score_1 = model.score(X_test_1, y_test_1)

Quick Debug Tip for Future #

To avoid this kind of mismatch issue down the line, add quick shape checks after loading your data:

print(f"Test features shape: {X_test_1.shape}")
print(f"Test labels shape: {y_test_1.shape}")

If the first number (sample count) doesn't match between the two, you'll immediately know there's a problem with how you're loading your labels.

内容的提问来源于stack exchange，提问作者AnDeep