社会システムモデルの多目標最適化について : 宮城ダイナミックモデルを事例として

概要

論文の詳細を見る
We know by experience the difficulty of managing to control the several parameters to lead DM (decision maker) into the satisfactory solution.We think it is due to too many parameter values to be tried and poor informations about the multi-objective criterions. In this point of view this study is focused on(1) computerizing the trial and error process, and(2) informing DM of his utility function which consists of multi-objectives through the learning processOur approach is a kind of goal programming method, where “goal” is revised every iteration phase and approaches to DMs preferential solution. DM is required only to choose his minimum acceptable value of each objective function, compared with trade off informations between the objective functions based on Geoffrion [3]. In the latter approach we felt the difficulty of the trade off estimation.This method is simpler and more heuristic than other interactive methods and is easy for DM to use it. Only one thing which DM should care for is to choose the parameter values so as to include DMs preferential solution.We have applied it to symplified Miyagi Dynamics Model and succeede.If objective functions are clearly hierarchically ordered, DM can set the preemptive priorities in this method. He learns which targets are satisfied in order of priority.
日本地域学会の論文