面向城市规划决策的地理国情数据挖掘——以北京市城市规模模拟为例

Mining National Geographic Census for Urban Planning Implications:A Case Geo-Simulation of Urban Size in Beijing

随着地理国情普查和监测的推进,信息量激增,数据挖掘方法和社会化应用工具不断进步,为解决规划决策中的资源分配率低、不确定性高等问题提供了得天独厚的优势。在北京市疏解非首都功能、推进京津冀地区协同发展的大背景下,以地理国情监测数据为基础,以复杂地理计算思想为指导,提出了一种基于集成学习的自适应元胞自动机模拟模型并应用于北京市城市规模演变的模拟。通过对比2015年北京市的预测结果与真实数据发现,相较于传统基于经验统计模型的元胞自动机模型,本文提出的模型预测精度更高。最后本文以2007年和2015年的北京市地理国情监测数据为模拟和训练数据,在假定当前社会经济发展速度不变的情况下,对2023年北京市城市规模作出预测,并找到北京市城市规模的演变规律,识别未来潜在的城市扩展热点区域,最终为城市规划的决策者提供"疏解"北京市的数据支持和现实路径。

With the development of the census and monitoring of geographical conditions,information content increases sharply.Progress of data mining methods and social application tools provides a unique advantage for solving the problem of low resource allocation rate and high uncertainty in planning decision-making.Against the background that Beijing municipalpolicy to easynon-capital function and promote synergetic development of Beijing-Tianjin-Hebei region,based on geographical conditions monitoring data,we proposed a new self-adaptive cellular automaton based on ensemble learning( EL-CA) model guided by complex geocomputing to the simulation ofcity scale evolvement in Beijing. Com-parison of prediction and real data in Beijing in 2015 proves that prediction of EL-CA model significantly out performs those of the traditional CA models based on empirical statistics. Finally we employed geographical conditions monitoring data of Beijing in 2007 and 2015 as training data,under the assumption that socioeconomic development remains unchanged,making predictions of city scalein 2023,trying to discover the rule of evolvement,identify potential hotspots of urban sprawl in the future and finally provide data supporting and practical path to in policy-making of"easing"Beijing.