[3E5-GS-2] Machine learning: Explainable AI (2)

[3E5-GS-2-03] Rule Extraction from Decision Tree Ensembles by Answer Set Programming

〇Akihiro Takemura^1,3, Katsumi Inoue^2,1 (1. SOKENDAI (The Graduate University for Advanced Studies), 2. National Institute of Informatics, 3. INTAGE Inc.)

Keywords:Answer Set Programming, Machine Learning, Knowledge Representation

Interpretability of trained models is one of the hotly discussed topics in the field of machine learning. In this work, we propose a method to extract rules from decision tree ensembles using answer set programming (ASP). We extract the rule sets in a declarative manner, by applying modified data mining methods encoded in ASP to the decision tree structure. As such, even if there are multiple types of rule sets that we would like to extract, our implementation can be used without too much modification. We also show some preliminary results obtained by applying our method to public datasets.