Encoding Random Forests with SAT for explainability

Advanced Artificial Intelligence course final project

The purpose of this project is to encode random forest models with SAT as explained in On Explaining Random Forests with SAT paper by Yacine Izza and Joao Marques-Silva with the final aim of explain model predictions.

Contents

docs

• documentation : examples of utilization of the implemented functions

src

• data : files related to datasets on which random forest model has been trained and tested

• lib : libraries used or created

• encode_rf_utils.py : implemented function to encode a Scikit learn random forest model with SAT

• mock_model.py : contains classes to instantiate a mock random forest model to be used in tests

• model : files related to the random forest model

• create_random_forest.py : python script to generate a random forest model on the Iris dataset as described in the reference paper

References

To generate the random forest model Scikit learn library has been used.
To encode and manipulate boolean and pseudo-boolean formulas PySAT library has been used.
The original work: [Yacine Izza, Joao Marques-Silva. 2021. On Explaining Random Forests with SAT]

@inproceedings{ims-ijcai21,
  author       = {Yacine Izza and
                  Jo{\~{a}}o Marques{-}Silva},
  editor       = {Zhi{-}Hua Zhou},
  title        = {On Explaining Random Forests with {SAT}},
  booktitle    = {Proceedings of the Thirtieth International Joint Conference on Artificial
                  Intelligence, {IJCAI} 2021, Virtual Event / Montreal, Canada, 19-27
                  August 2021},
  pages        = {2584--2591},
  publisher    = {ijcai.org},
  year         = {2021},
  url          = {https://doi.org/10.24963/ijcai.2021/356},
  doi          = {10.24963/ijcai.2021/356}
}

And the github repo