Hi,
Thank you for your very helpful implementation codes.

I got some questions about 'LSTM_CNN_Multi.py'.

According to paper, the proposed model consists of 3 multiple pipeline.
However, I can not find the multiple pipeline in the code file...

Is there any something that I miss..?

Hi, I came across your repo

in the following section of your code you are providing information about the future to the model:

for i in range(0 , len(stock) - window_size -1 , 1): first = stock.iloc[i, 4] temp = [] temp2 = [] for j in range(window_size): temp.append((stock.iloc[i + j, 4] - first) / first) # for j in range(week): temp2.append((stock.iloc[i +window_size, 4] - first) / first) # X.append(np.array(stock.iloc[i:i+window_size,4]).reshape(50,1)) # Y.append(np.array(stock.iloc[i+window_size,4]).reshape(1,1)) # print(stock2.iloc[i:i+window_size,4]) X.append(np.array(temp).reshape(50, 1)) Y.append(np.array(temp2).reshape(1,1)))

for i = 0 temp gets the first 50 close values of the csv
these are beeing saved in the X variable
but you also save the 50th close value in temp2 and therefore in Y
This means that the model already has the value we want to predict in Y

Also the accuracy the model has after training for only 1 epoch is to good to be true