A machine learning-enhanced fuzzy decision-making model for blockchain platform selection in healthcare systems

Abstract

The healthcare sector is experiencing unprecedented growth in the volume and complexity of its data, driven by the increasing digitization of patient records, the interconnectedness of devices, and the involvement of diverse stakeholders. While blockchain technology offers significant potential to address critical issues of security, transparency, integrity, and accessibility, selecting an appropriate blockchain model for healthcare remains challenging due to the multiplicity of technical, operational, and regulatory considerations. This study proposes a novel methodological framework that systematically integrates Machine Learning (ML) and Multi-Criteria De cision-Making (MCDM) methods to enhance the stability and reliability of the decision process. First, Agglom erative Hierarchical Clustering (AHC) is applied to eliminate redundancy and reduce dimensionality among the large set of decision criteria identified from the literature. Next, Proportional Picture Fuzzy Sets (PPFSs) are employed to more accurately capture the complexities of human judgment, including hesitation, disagreement, and partial agreement, providing a more detailed picture than traditional fuzzy methods. The relative importance of the criteria is determined using the PPFS-based Weights by ENvelope and SLOpe (WENSLO) approach, which ensures data-driven and objective weight assignment under uncertainty. Subsequently, blockchain alternatives are ranked through the PPFS-based Complex Proportional Assessment (COPRAS) method, which proportionally incorporates both favorable and unfavorable attributes. The findings indicate that private (permissioned) blockchain architectures emerge as the most suitable option for healthcare settings, largely because they perform better on critical criteria such as data decentralization, efficiency, and accessibility. Finally, a sensitivity analysis and comparative assessment are performed to check the robustness and stability of the developed model.