使用Keras的classifier.fit()方法时出现输入形状不匹配错误求助

Hey there, let's get this shape mismatch sorted out quickly. That error message spells out exactly what's going wrong: your model expects each input sample to have 11 features (shape (11,)), but right now your x_train is feeding it samples with 16934 features each (shape (16934,)). Here's how to fix it:

Step 1: Diagnose the Current Shape of x_train #

First, let's confirm what we're working with. Run this line to check the shape of your training data:

print(x_train.shape)

You'll likely see one of two outputs, and we'll handle each case below.

Case 1: Output is (16934,) #

This means x_train is a 1D array—probably because you accidentally flattened your 2D dataset (where each row should be a sample with 11 features). To fix this, reshape it into a 2D array where each row is a sample with 11 features:

# Reshape to (number_of_samples, 11) - the -1 lets Keras calculate the sample count automatically
x_train = x_train.reshape(-1, 11)

Double-check the shape again after this step—it should now show something like (16934, 11) (assuming 16934 is your total number of samples).

Case 2: Output is (11, 16934) #

This means you've got your dimensions flipped: your data is arranged as (number_of_features, number_of_samples) instead of the Keras-expected (number_of_samples, number_of_features). Fix this by transposing the array:

x_train = x_train.T

After transposing, the shape should be (16934, 11), which matches what your model expects.

Step 2: Verify and Retrain #

Once you've adjusted the shape, run print(x_train.shape) again to make sure it's (N, 11) where N is your number of training samples. Then try running your classifier.fit() call again—it should work without the shape error now.

A quick side note: Make sure your y_train is also in the correct shape (either (N,) or (N, 1) for binary classification) to avoid any additional mismatches.

内容的提问来源于stack exchange，提问作者Thinker