"Automation of Machine Learning Pipeline Design by an Ontology as an Integrative Meta-Learning Model"

This article considers the problem of automating the design of machine learning (ML) pipelines. Methods for automating the design of ML pipelines were analyzed. Based on the analysis performed an ontology would be promising for solving the above problem. A method for automating the design of ML pipelines based on ontological engineering was proposed. An ML ontology aimed at constructing pipelines was created. An application for automated construction of pipelines based on the created ontology was developed. The effectiveness of the developed solution has been assessed experimentally, as compared with state-of-the-art automated pipeline construction tools named TPOT. The solution presented not only appears to be more efficient in terms of the quality of the result obtained in the minimum required time, but is also comparable to the above tool regardless of the time of running.