住宅地价特征识别与规划增强的机器学习诊断  

Machine Learning Diagnosis of Residential Land Price and Its Planning Enlightenment

作　　者：彭坤焘 赵绮琪 PENG Kuntao;ZHAO Qiqi

机构地区：[1]重庆大学建筑城规学院山地城镇建设与新技术教育部重点实验室 [2]重庆大学建筑城规学院

出　　处：《上海城市规划》2022年第5期33-39,共7页Shanghai Urban Planning Review

基　　金：国家自然科学基金“技术—经济范式变迁与产业空间形态演变的关联性及规划策略研究”(编号51608062);重庆市自然科学基金“重庆主城住宅地价特征识别与规划增强的机器学习诊断”(编号CSTB2022NSCQ-MSX0359)资助。

摘　　要：住宅地价一定程度上映射了城市效用,充分映射时可称为影子地价。现实中,成交地价是不完全信息下有限理性的博弈结果,证据之一是出让方式会影响价格,拍卖方式相比零溢价挂牌通常存在着较高溢价。假如住宅地价有合理稳定的估值,那么成交地价偏离影子地价将会带来人地关系、公共财政、城市形态等方面的问题,需进行公共干预。为了准确判断规划增强必要性,可以提炼和采集影响住宅地价的多维特征变量,采用交叉运用回归与分类的机器学习方法进行诊断。研究采集了2020年以来重庆主城住宅用地出让的公开数据,运用LGBM模型开展诊断和预测,并从土地出让机制、异质发展、效用收敛3个方面提出规划增强策略的积极作为的建议。The residential land price reflects the urban utility to a certain extent,and can be called the shadow land price when it is fully mapped.In reality,the actual land price is the result of the game of bounded rationality under incomplete information.One of the pieces of evidence is that the transfer method will affect the price.Compared with the zero premium listing,the auction method usually has a higher premium.If the residential land price has a reasonable and stable valuation,the deviation of the actual land price from the shadow land price will bring about problems in the man-land relationship,public finance,and urban form,which requires public intervention.In order to accurately judge the necessity of planning enhancement,it is necessary to extract and collect the multidimensional characteristic variables that affect the housing land price,and use the artificial intelligence method of cross-use regression and classification for diagnosis.Based on the public data on residential land transfer in the main city of Chongqing since 2020,the research adopts the LGBM model to explore the diagnostic method.From three aspects of the land transfer mechanism,urban heterogeneous development and urban utility convergence,this paper puts forward positive actions of planning enhancement strategy.

关 键 词：人工智能 LGBM模型 住宅地价 空间公平 

分 类 号：TU984[建筑科学—城市规划与设计]