How to get the feature importance of ranklib generated random forests model?

Using ranklib's learning to rank random forests generates an xml-like model. Ranklib has a tool that provides features' frequency which cannot necessarily be considered as feature importance.

How can I get the Gini feature importance or Gini index of random forests generated by ranklib? How to parse the tree generated?

Found in the Sourceforge discussion forum that you need to parse the model file yourself.

I, personally, had a lot of struggles to get the Gini importance of features from a ranklib random forest and finally succeeded. Here I share the Github repository I made to solve the problem.

You can do it by running this command (use python3):

python Gini.py <num_features> <path_to\training_data> <path_to\RF_model> <trees directory> <output_file>

Please see the repository for more details on how to do it.