Quantum-Driven Predictive Business Models for Optimizing Supply Chain Management and Strategic Planning
Annotatsiya
Classical optimization methods handle supply chain planning problems poorly at enterprise scale because their computational complexity grows exponentially with instance size, forcing practitioners to accept suboptimal solutions within fixed time budgets. Quantum computing offers a complementary capability: QAOA-based combinatorial optimization solves routing, inventory-allocation, and disruption-scenario enumeration problems with efficiency advantages that classical MILP cannot match, while remaining weaker than classical ML at the demand-forecasting tasks that supply those optimizations. This paper proposes a Hybrid Quantum-ML framework combining QAOA with an LSTM ensemble forecasting layer, applied to 48,000 strategic planning decisions across 18 enterprise deployments (2,400,000 supply transactions, twelve-month horizon; 75/25 chronological split: 36,000 train / 12,000 test). The hybrid achieves optimization accuracy of 0.884 against expert-validated planning benchmarks, outperforming classical MILP (0.684; +29.2% [(0.884-0.684)/0.684]), ML-only LSTM ensemble (0.764; +15.7%), and quantum annealing-only (0.804; +10.0%). Deployment outcomes include supply chain cost reduction of 32.6%, planning horizon extension from 14 to 84 days (+500%), and mean disruption-recovery time reduction from 14.4 to 7.4 days (−48.6%). Sub-domain accuracy spans demand forecasting$(\mathbf{r}=\mathbf{0. 8 8 4})$, inventory optimization (0.864), disruption prediction (0.824), and route optimization (0.804). Quantum decoherence under NISQ hardware degrades accuracy to 0.784 (−11.3%) and small-instance problems below the quantum-advantage threshold to 0.724 (−18.1%), each with a deployed mitigation documented in Section V.
Hali tarjima qilinmagan