The Generation Method of Simulation Scenario Sample Space Based on Sensitivity Analysis of Meta-model

Abstract

To ensure the feasibility and effectiveness of exploratory simulation experiments, it is necessary to take the simulation scenario sample space with acceptable scale and typical representative as input. In this paper, a method of generating simulation scenario sample space combining qualitative and quantitative analysis is proposed. This method constructs a machine learning meta-model based on simulation pre-experiment, and screens the key experimental factors based on sensitivity analysis of meta-model to determine the factor levels. Finally, the space is sampled and compressed to complete the generation of the hypothetical sample space.