Ontology-Driven Machine Learning: a Review of Applications in Healthcare, Finance, Natural Language Processing, and Image Analysis

EasyChair Preprint no. 10430

Abstract

Ontology-driven machine learning (ODML) is an emerging approach that integrates domain-specific ontologies with machine learning methods to improve the accuracy, interpretability, and explainability of predictive models. In this review, we synthesized the findings from studies that applied ODML in different domains, including healthcare, finance, natural language processing, and image analysis. The results demonstrated the potential of ODML in improving the performance of machine learning models for predicting falls, hospital readmissions, credit risk, stock prices, sentiment analysis, text classification, breast cancer histology, and liver image classification. The study highlighted the importance of domain-specific ontologies in capturing the domain knowledge and improving the performance of machine learning models. However, the quality of evidence varied across the studies, and publication bias may be present. Future studies should aim to address these limitations and develop standardized approaches for ontology development and integration with machine learning methods. Overall, ODML is a promising approach for improving the accuracy, interpretability, and explainability of machine learning models in different domains.

Keyphrases: Applications, domains, machine learning, ontology-driven

@Booklet{EasyChair:10430,
  author = {Admas Abtew and Dawit Demissie and Kula Kekeba},
  title = {Ontology-Driven Machine Learning: a Review of Applications in Healthcare, Finance, Natural Language Processing, and Image Analysis},
  howpublished = {EasyChair Preprint no. 10430},

  year = {EasyChair, 2023}}