ABSTRACT
Objective:To optimize the dispatch of pre-hospital emergency resources and address the assessment challenge of ambulance demand, a pre-hospital emergency demand prediction model based on multi-model fusion was constructed.Methods:The retrospective study design method was adopted, and historical pre-hospital emergency dispatch records and corresponding weather data were extracted. Three types of primary learners were trained by 5-fold cross-validation, and the training results of the primary learners were fused by Stacking. The fusion results were input into the secondary learner as new features, and the final prediction results of ambulance demand were obtained by the secondary learner.Results:By comparison experiments, results showed that the multi-model fusion prediction model based on Stacking was superior to the single model in both mean absolute error and root mean square error, indicating that the model could predict ambulance demand more accurately.Conclusion:The pre-hospital emergency demand prediction model based on multi-model fusion could improve the accuracy and generalization ability of ambulance demand prediction by using historical emergency data and weather data, and provide strong support for the optimization of pre-hospital emergency resources.