摘要：Day-ahead emergency blood demand (EBD) forecasts after earthquakes play a key role in the success emergency blood operations. Demand uncertainties and limited data pose great challenges for accurate EBD forecasts. Given such challenges, we build an adaptive evolutionary support vector regression (AESVR) model to forecast daily EBD. We also build seven other potential models for comparison. Real-word case and data are used to evaluate these models. Computational results demonstrate that the AESVR generates a favorable accuracy. Ranking orders regarding accuracy and complexity of all models can provide guidance for decision-makers and practitioners to make tradeoffs according to their own preferences.
