问题
I have developed a RNN system to predict the stock market close price based on several features. I removed the close price from the X and moved it to Y. I also made a shift on Y and dropna() to make sure that all features of day X(t) is equals to Y(t+future_days). Thus means that the features for today will predict the close price of today plus future_days.
From the images bellow, the first future_days = 20 and the second future_days = 5. The prediction is shifted the number of future_days in each image. I am ploting the predictions and close, where close is equals to Y_test, which is also shifted future_days. Thus, I was expecting that Y_Test would be the same of prediction and the data would not be shifted in time.
来源：https://stackoverflow.com/questions/60150501/lstm-time-series-shifted-predictions-on-stock-market-close-price