如何使用 keras 将数组重塑为 3d

Question

I have 1260 rows and 2 columns that represent a feature per column (ie 'Open','High') for stock predictions. I need to reshape this array into the 3D array for input into Nerual Network using LSTM.

This is my code so far:

X_train, y_train = np.array(X_train), np.array(y_train)
X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], 2))

The last line of the code I am confused about what it is telling me. It's meant to put it into a 3D array but it seems like a 2D array to me.

How can I feed the network an array (or vector of Open values) and an array of High Values.

Answer 1

I assume X_train.shape is a 2D tensor, if you want to make it 3D you can use np.expand_dims

Example

arr = np.random.randint(0,5,(3,3))

array([[3, 4, 1],
       [1, 0, 1],
       [4, 4, 2]])

arr.shape = (3,3)

reshape_arr = np.np.expand_dims(arr, -1)
reshape_arr.shape = (3,3,1)