Predictive Control of Clustered Closed-Loop Supply Chains based on Multimodal Support Vector Machines

To address inventory fluctuation and allocation coordination challenges faced by clustered closed-loop supply chains under demand uncertainty, this paper proposes an inventory allocation optimization method integrating demand forecasting with model predictive control (MPC) based on multimodal support vector regression. This approach utilizes numerical features such as historical sales data alongside sentiment features extracted from product reviews to establish a multimodal demand forecasting model. The forecast results are then embedded as exogenous inputs within an MPC control framework, enabling rolling optimization and coordinated allocation of inventory across multiple chains. Simulation comparisons with both basic MPC and unimodal SVR+MPC methods validate the improvements achieved by the multimodal SVR+MPC approach in demand forecasting accuracy, inventory fluctuation suppression, and system operational stability.