We use the following function to create the training and test sequences that we will use to train and test our networks. The function takes a set of time series stock prices, and organizes them into segments of n consecutive values in a given sequence. The key difference will be that the label for each training sequence will correspond to the stock price four timesteps into the future! This is quite different from what we did with the moving average methods, as they were only able to predict the stock price one timestep in advance. So, we generate our sequences of data so that our model is trained to foresee the stock price four time steps ahead.
We define a look_back value, which refers to the number of stock prices we keep in a given observation. In our case, we are actually allowing the network to look_back at the past 7 price values, before...
