NumPy重塑数组时触发IndexError: tuple index out of range问题求助

Let's walk through why this error is happening and how to fix it for your time series data prep.

What's Causing the Error? #

The root issue is in how you're slicing your scaled_data to build x_train:

x_train.append(scaled_data[prediction_days-i:,0])

When i starts at prediction_days (100) and increments, prediction_days - i becomes a non-positive number:

• For i=100, it's 0 → you're taking the entire first column of scaled_data (length equal to the full dataset)
• For i=101, it's -1 → you're taking only the last 1 element
• For i=102, it's -2 → you're taking the last 2 elements, and so on

This means every entry in x_train has a different length. When you convert this list to a NumPy array with np.array(x_train), you don't get a 2D array like you expect—you get a 1D object array (shape looks like (N,), where N is the number of samples).

Since this array only has one dimension, trying to access x_train.shape[1] (the second dimension) throws the IndexError: tuple index out of range—there is no index 1 in a 1-element shape tuple.

How to Fix It #

You're trying to use the previous prediction_days days of data to predict the current day's value, so you need to reverse the slicing logic to get a consistent window of 100 days per sample.

Update your loop code to this:

for i in range(prediction_days, len(scaled_data)): 
    # Grab the 100 days BEFORE the current index i
    x_train.append(scaled_data[i-prediction_days:i, 0]) 
    y_train.append(scaled_data[i, 0]) 

After this change:

1. Each entry in x_train will be a 100-element array (matching your prediction_days value)
2. Converting to np.array(x_train) will give you a 2D array with shape (len(scaled_data)-100, 100)
3. The subsequent reshape call will work as intended, converting the 2D array into the 3D format required for LSTM models: (number of samples, time steps, features)

Quick Verification Step #

Add a print statement after converting to arrays to confirm the shape is correct:

x_train, y_train = np.array(x_train), np.array(y_train)
print("x_train shape:", x_train.shape)  # Should output (X, 100) where X is your sample count

内容的提问来源于stack exchange，提问作者legar