Parameter optimization of simulation models for material supply in an emergent disaster based on support vector machine

Modeling material supply in emergent disasters has become an effective means to foster emergency relief, for which Agent-DEVS (discrete event system specification) model parameter optimization is important. Based on fundamental support vector machine (SVM) principles, a parameter optimization flow for an Agent-DEVS model is put forward, and a parameter optimization model for the supply task parameter in simulation models for material supply in an emergent disaster is established. Some key techniques, including data extraction and preprocessing, kernel function selection and SVM model parameter preferences, are analyzed, and the comparison with the back-propagation neural network is examined. A simulation test shows that SVM has strong learning and fitting capabilities and weak dependence on samples. It enhances the dynamics of Agent-DEVS models. The self-learning ability significantly improves model intelligence and the optimized parameters provide models with more elaborate descriptive abilities.