Human in the Loop Fuzzy Pattern Tree Evolution

Fuzzy pattern trees evolved using grammatical evolution, a system we call Fuzzy Grammatical Evolution, are shown to be a robust Explainable Artificial Intelligence technique. Experimental results show Fuzzy Grammatical Evolution achieves competitive results when compared against SVM, Random Forest and Logistic Regression on a set of real world benchmark problems. Fuzzy Grammatical Evolution allows for human input throughout the evolutionary process. Regularization methods and double tournament selection were investigated to determine what method was most successful at finding smaller, more interpretable models. A domain expert was recruited to investigate the interpretability of the models found and to give a confidence score for each model. This expert successfully identified overfit models, validating that Fuzzy Grammatical Evolution can be regarded a powerful Explainable Artificial Intelligence technique.