Figure 2.

Experiment pipeline for the learning models development, with the isolated implementation of the RF algorithm (Experiment 1), the implementation of dimensionality reduction (DR) and feature engineering (FE) alongside the tuned RF (Experiments 2 and 3, respectively), and the final application of oversampling (O) with the previous techniques (Experiments 4 and 5). Red squares indicate that hyper-parameter tuning is being performed to a specific technique while red arrows indicate the use of a technique tuned in a former experiment.