MonoLoop:CA城市模型状态转换规则获取的一种方法

状态转换规则是元胞自动机(Cellular Automata,CA)的核心,如何获取并建立CA的状态转换规则是构建CA模型的关键。邻域作用是CA能够模拟复杂物理现象的核心驱动力,而在已有的用于城市增长模拟的CA城市模型中,因为邻域作用在模拟的过程中为时间动态的变量,其系数很难通过常用的Logistic回归方法识别,致使已有的CA城市模型的状态转换规则中,往往仅通过Logistic回归获取邻域作用之外的空间变量的模型参数,而邻域作用的参数通常采用主观赋值的方法。本文提出了CA城市模型的多指标评价(Multi-Criteria Evalua-tion,MCE)形式状态转换规则获取的一种新方法 MonoLoop,并针对北京市域1976～2006年的城市增长开展了该方法的实验。基于这种方法,一方面利用历史数据可以建立更为客观的状态转换规则;另一方面也可以大大降低模型参数识别的时间。

MonoLoop: A New Approach to Retrieves Status Transition Rule of CA Urban Model

Status transition rule is the core component of cellular automata,and retrieving status transition rule for CA is the key issue to develop a CA based model.Neighbor influence is the basic driving force for simulating complex physical phenomenon.However,the coefficient of neighbor cannot be identified via traditional logistic regression for its temporal dynamic character.In current CA urban models,coefficient of neighbor is usually set in subjective way.We bring forward an approach named "MonoLoop" to retrieve MCE based status transition rule of CA urban model,and conduct an experiment,using historical urban growth dataset of Beijing metropolitan area in 1976-2006,to test MonoLoop method.It has proved that MonoLoop is capable to identify more objective and accurate status transition rule with less time consuming.