The following exercise is to train a Recurrent Neural Network to follow the directions taken by our predictions, rather than the closeness of their values to the real stock price. We want to check if our predictions follow the same directions as the real stock price.
Here, the model is created using Stacked LSTM units.
The training dataset consists of Google Stock Prices for the past 5 years including only the working days. The test set or the prediction set consists of stock prices for the month of January 2017.