A search method of inputs to target outputs on qualitative and quantitative hybrid simulation using neighbor selection by sensitivity analysis

This paper addresses a search problem of inputs to target outputs on qualitative and quantitative hybrid simulation. The hybrid simulation can simulate qualitative factors in business quantitatively using Monte Carlo simulation that executes simulations repeatedly using random numbers. Because of using not equation but random numbers, it is difficult to search inputs deriving target outputs that a business manager expects. The general approach, an iterative search method of inputs, takes much time in case of large-scale simulation. So, we propose a search method using neighbor selection by sensitivity analysis. Until the target outputs are derived, our method repeats generating neighbors of a certain inputs and selecting a neighbor that tends to derive target outputs. The neighbor is selected by sensitivity analysis, which is based on a distribution distance defined by target outputs and simulated outputs in terms of averages and variances of distributions. By applying our method to a qualitative and quantitative model, it is confirmed that the computational time is decreased by our method.