Data Preparation – Creating the Past
Let's now implement in practice a demand prediction application, using the time series for cluster 26. Again, we will have two separate workflows: one to train the LSTM-based RNN and one to deploy it in production. Both applications will include a data preparation phase, which must be exactly the same for both. In this section, we will go through this data preparation phase.
Dealing with time series, the data preparation steps are slightly different from what is implemented in other classification or clustering applications. Let's go through these steps:
- Data loading: Read from the file the time series of the average hourly used energy for the 30 identified clusters and the corresponding times.
- Date and time standardization: Time is usually read as a string from the file. To make sure that it is processed appropriately, it is best practice to transform it into a Date&Time object. A number of nodes are available...
