基于集成学习和元胞自动机的城市地理模拟

Urban Geosimulation Based on Ensemble Learning and Cellular Automata

针对常用城市地理模拟系统中元胞自动机转换规则获取算法的局限性,提出基于集成学习的元胞自动机,并将其应用于城市建设用地的动态模拟.以决策树作为弱分类器,应用集成学习和元胞自动机,模拟了东莞市2001—2005年的建设用地时空格局.精度评估的结果表明,经集成学习后的决策树比单个决策树对城市建设用地动态的模拟精度更高,算法泛化能力更好.

In order to alleviate the limitation of obtaining transformation rules in GIS using cellular automata, a cel- lular automata based on ensemble learning is proposed for simulating urban dynamic geosimulation. Decision tree is used as weak classifier in the ensemble learning and cellular automata to simulate the urban spatio-temporal dynam- ics in Dongguan from 2001 to 2005. The accuracy results show that the simulation of ensemble learning is better than using decision tree alone for urban dynamic geosimulation; The new method can obtain better generalization a- bility.