Machine learning in humanitarian logistics and supply chain management: application domains and quality assurance

Purpose Despite the growing integration of machine learning (ML) in humanitarian supply chain (HSC), herein HSCML, no prior review has established a systematic application domain taxonomy or examined model quality assurance, limiting sustainable adoption. This study aims to address three gaps: the absence of a systematic taxonomy for HSCML application domains, the lack of frameworks for HSCML data and model quality and the disconnect between technical development and operational deployability. Design/methodology/approach A sequential mixed-methods systematic literature review was used. In Phase 1, latent Dirichlet allocation topic modelling was applied to 302 HSCML journal papers to identify thematic clusters and derive a context–intervention–mechanism–outcome (CIMO) classification system. In Phase 2, a CIMO-guided systematic review analysed 93 HSCML model development studies across contextual, methodological and quality dimensions. Findings ML contributed to HSC through six domains: disaster consequence prediction, relief operations optimisation, real-time crisis intelligence, infrastructure monitoring and recovery, population mobility and evacuation and strategic HSC network planning. Research efforts are unevenly distributed across disaster types, management phases and regions, with fairness, explainability and uncertainty quantification systematically neglected across all domains. In response, three frameworks are established covering data quality dimensions, model quality dimensions and factors influencing model quality. Originality/value This is the first systematic examination of data and model quality in HSCML. It delivers a deployability-focused quality assurance framework and a data-centric research agenda addressing 14 knowledge gaps, providing practical references for rigorous HSCML development in line with humanitarian obligations.