First version refactory

In this release, we kept the same functionality of the first version; however, we applied several improvements in the code architecture to allow its extension. Additionally, we include the support of GRU models.

First stable version

This version includes:

• Training of embedding layers for activities and roles encoding.
• Preprocessing of event-logs using n-grams and scaling techniques for creating sequences examples to train LSTM models.
• Training scripts of different LSTM architectures.
• Post-processing options about the selection of next activity, by using Arg. Max or RandomChoise following a PDF.
• Prediction scripts for the next event of a process, the case continuation, and the generation of a full event-log from zero size prefix.