Summary
Arriving at the end of this chapter, we have demonstrated how to expand a prediction model from a one-dimensional time series (univariate) to a multi-dimensional time series (multivariate). We expanded the input from the past values of a single time series to include the past values of all 30 time series in the energy consumption dataset and learned how to build a model predicting the next value in one of the selected time series.
We approached the problem in steps. First, we trained a fully connected feedforward neural network to predict the next value in one time series based on the past values of all time series. Then, we trained a fully connected feedforward neural network to predict the next values in all 30 time series based on the past values of all 30 time series in the energy consumption dataset.
Finally, we observed that the more complex the problem and the model, the higher the computational load, and the longer the execution times, especially during the training...
