International University of Management Stock Price Prediction Program

See attached images <<<<<<<<<<<<<<<<

Predict Stock Market in the Future
This exercise is used to replace the constraint-time exam of the BAAI course. You are going to solve
a problem of predicting the future stock market with the LSTM model in Keras.
O
Frist, we need the dataset of stock market, e.g., S&P 500 index from investing.com. This dataset
includes 3,590 rows and 7 feature variables. Each row corresponds to a day, and includes the
variables:
Date: Time Variable
Price: S&P 500 index
Open: Open price of this date
High: The highest price of this date
Low: The lowest price of this date
Volume: An empty column
Change %: Compared to the previous date
.
.
.
Please modify the codes from Lecture 11 and report the results of Mean Absolute Error and Root
Mean Square Error on both Training dataset and Test dataset. Please note
You want to predict the price after 30 days with the previous prices of 60 days. (Note it is
different from the prediction setup we practiced in Lecture 11. The difference is illustrated in
the figure below.)
You use previous 60 day prices to build model
Quiz 2:
60 day ago
Today
30 day later
Lecture 11:
60 day ago
Price that
you predict
Today
Tomorrow
You use previous 60 day prices to build model
.
0
You split the dataset with the ratio 7-to-3 for the training set and the test set in chronological
order.
You use 100 LSTM nodes in the layer.
You optimize the model for 50 epochs.
You make a plot of the Mean Absolute Error vs. Epochs for training and test data.
You make a plot for comparing the real S&P 500 index and the predicted price from your
model.
You need to explain what can be the reason that you get inferior prediction result for 30 days
later compared to the prediction for tomorrow.
.
You may add more layers in your LSTM model to achieve better results after reporting the required
tasks.