模型验证——基于主体建模的方法论问题

基于主体的(agent-based, AB)建模已成为目前社会经济系统复杂性分析的重要方法, 但由于缺乏有效的模型验证导致AB 建模方法还未能得到主流研究者的认同. 本文比较了AB建模和基于方程(equation-based, EB)建模在建模过程上的区别, 从结果验证、过程验证和模型对接研究三个方面回顾了AB模型验证的主要方法论问题和相关研究进展, 指出建模工具标准化、一致性判断、灵敏度分析和参数空间压缩技术是未来AB 模型验证的主要方向.

Model validation — Methodological problems in agent-based modeling

Agent-based modeling has become an important method for analyzing the complexity in socio-economic system. But due to lack of effective model validation, this method has not been accepted by the mainstream researchers. The paper compares agent-based modeling with equation-based modeling in aspect of modeling procedure, and reviews the main methodological problems and relative research advances in validation of agent-based models from three aspects: outcome validating, procedure validation and model docking. We argue that standardization of agent-based modeling tools, consistency judgment, sensitivity analysis, and parameter space reduction technique will be the promising research directions in validating agent-based models.