Towards hybrid self-learning ontologies: A new Python module for closed-loop integration of decision trees and OWL

Abstract

Ontologies are a valuable tool for organising and representing knowledge. Decision trees (DTs) are a machine learning (ML) technique that can be used to learn from data. The integration of DTs and ontologies has the potential to improve the performance of both technologies. This paper proposes a novel approach to integrating DTs and ontologies. For that purpose, a new Python module is developed and validated to demonstrate this approach. The dt2swrl module has been evaluated against two research questions (RQs). RQ1 asked whether existing Python packages are capable of integrating DT rules with ontologies. The investigation showed that none of the existing packages can achieve the defined goals. RQ2 asked whether a generic module can be developed to automatically integrate DT rules with ontologies. The dt2swrl module has been developed to address this gap and it has been shown to be effective in achieving the desired goals. It can be used to develop hybrid ontologies that seamlessly integrate expert-based and data-based rules, and self-learning ontologies that can automatically maintain their rules based on new data. The paper concludes by discussing the limitations of the dt2swrl module and the implications of the research for future work.