Vol. 1 No. 2 (2025): Journal of Computing and Engineering - JCE

Black-Box Machine Learning versus White-Box Methods: A Mapping Literature Review

Mohammed Ghazi Naji Al-Nussairi ¹

¹Department of Computer Science, University of Oviedo, Oviedo, Spain, UO298481@uniovi.es, ORCID 0009-0005-7568-3541

The rapid proliferation of machine learning (ML) and deep learning (DL) methods across scientific and industrial domains has intensified the debate between predictive performance and model interpretability. Black-box models, including deep neural networks, ensemble methods, and support vector machines, offer state-of-the-art accuracy but operate with limited transparency. White-box models, encompassing linear programming, interpretable ML, and rule-based systems, prioritize explainability, auditability, and constraint satisfaction. This systematic literature review synthesizes findings from 50 peer-reviewed articles published between 2015 and 2025, retrieved from IEEE Xplore, Scopus, and Web of Science. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework, this review compares black-box and white-box approaches across six dimensions: predictive performance, interpretability, computational efficiency, scalability, robustness, and domain applicability. Results reveal a persistent accuracy–interpretability trade-off, an emerging convergence through explainable artificial intelligence (XAI) methods, and growing hybrid architectures that blend both paradigms. This review identifies critical research gaps and provides a structured roadmap for future investigation.

Keywords: Black-box models, deep learning, explainable AI, interpretable machine learning, machine learning, optimization, literature review, white-box models, XAI.