SIMPLIFYING RANDOM FORESTS THROUGH POST-HOC RULE EXTRACTION

Despite the high accuracy of black-box models, a significant challenge remains: their decision-making processes are often too complex for humans to easily understand. In response, there has been a renewed attention to explainable and interpretable artificial intelligence, a field dedicated to making the decision-making processes of models more understandable. Building upon prior work and using the Random Forest model as a basis, we create a rule extraction framework which seeks to produce a more understandable model that retains predictive performance. Through the use of post-hoc rule extraction methods, we extract rules from the original ensemble, reduce the size of the ruleset, and thus improve the overall explainability.