Our idea is simple at it's core. Use Mutual Information to assist feature selection in random forests judiciously in order to retain the robustness inherent to Random Forests and boost it's performance.
The dependencies for this project are rather minimal, including,
- Numpy
- Matplotlib
- Scikit-learn 0.19+
The design of FSRF is identical to that of sklearn routines.
To be put in repo soon
Check Test_foresight.ipynb
for the time being