We present GeoDepend-ML, a temporal, multiplex Ecore metamodel for representing geopolitical interdependence and influence with auditability. The model treats actors (states, blocs, enterprises, organizations), assets (resources, technologies, infrastructures, routes), and artifacts (e.g., treaties, licenses, sanctions) as first-class elements, and encodes typed relations—Dependency, Influence, Control, and Flow—across layers (resource, technology, finance, information, route, norms, diplomacy, security). Each relation carries quantitative attributes (typed Quantity with UCUM units), quality, duration, exclusivity, substitutability, reliability, cost, legal risk, and strategic sensitivity, together with provenance and uncertainty descriptors. We formalize temporal validity and add OCL constraints (e.g., license ⇒ technology) plus an analytic operation to compute an effective dependency weight. A compact multi-layer case study (critical materials, advanced lithography, maritime route) illustrates expressiveness, consistency checking, and sensitivity analysis for “what-if” scenarios. This evaluation is intentionally limited in scale to highlight semantic clarity and analytic behavior; integrating GeoDepend-ML with large geopolitical datasets such as ICEWS or GDELT is planned for future work. Intended applications include auditable dependency assessment, scenario-based risk analysis, leverage estimation, and decision support for governments, international organizations, and strategic sectors. By combining provenance-aware structure with uncertainty handling, GeoDepend-ML lays a reproducible foundation for auditable geopolitical decision support and downstream analytics (e.g., dependency indices, coercive leverage, and shock propagation.

Multiplexing, Uncertainty, Sensitivity analysis, Semantics, Lithography, Licenses, Security, Risk analysis, Reliability, Sustainable development, Ecore, EMF, international relations, knowledge graphs, multiplex networks, decision support, provenance, uncertainty