This project attempts to reproduce the procedure and results achieved by Kerschke et al. and Prager et al.. In addition, the feature normalization presented in Prager & Trautmann is taken into account.
- Install required packages (all listed in the first notebook cell)
- Inside a section execute the cells in order (the sections itself are independent of each other)
- In model training, the cells for training the models can also be executed independently of each other
- Sections Feature calculation and relERT conversion can be skipped to train the models as all features and relERT are already precomputed
- The results for the different models are also already available, so it is also possible to visualize only the results using the last section