Active learning metamodelling for survival rate analysis of simulated emergency medical systems

Emergency Medical Services (EMS) constitute a crucial pillar of today's cities by providing urgent medical responses to their citizens. Their study is often conducted via simulation, as the assessment of planning decisions is generally unfeasible in the existing systems. However, such models can become computationally expensive to run. Thus, metamodels can be used to approximate the simulation results. In this work, a simulation metamodelling strategy supported on an active learning scheme is proposed to analyse the survival rate of a simulated EMS. The exploration process is guided through a series of grids towards simulation input regions whose output results match a specific survival rate defined a priori. This provides an efficient way of exploring the search space by channelling the computational effort to the most important input values, supporting the advantages of these methodologies in the EMS field, where their application is still seldom to the best of our knowledge.