【作者】 李霞；

【导师】 邵春福；

【作者基本信息】 北京交通大学 ， 交通运输规划与管理， 2010， 博士

【摘要】 城市通勤交通与居住就业空间分布之间存在互动关系。国内外研究表明,把握城市中交通与居住就业空间关系的形成机理和影响因素,揭示其内在的规律,了解居民在进行居住空间选址时的行为规律和互动关系,对于制定切实可行的城市交通政策尤为重要。本论文以我国大城市为对象,从微观和宏观的角度,研究城市通勤交通与居住就业空间分布相互关系模型与方法,为城市交通规划的制定、交通政策分析以及城市空间结构调整提供理论基础。首先,归纳总结国内外关于城市交通与居住就业相互关系的研究历程、研究成果及未来研究方向,为论文的研究提供基础。第二,从宏观层面,分析城市通勤交通与居住就业空间分布的互动机理,从微观层面分别用网络广义极值理论和小波神经网络理论建立城市居民居住就业和交通方式联合选择模型,并对两类模型的解释能力和预测能力加以对比分析。第三,从宏观和微观角度研究轨道交通对居住就业和出行方式选择的影响,用交叉分层Logit模型对轨道站点周边居住地选择和出行方式选择进行建模,同时在样本数据有限情况下研究城市轨道交通沿线房地产价格的变化规律,提出基于支持向量机的回归预测模型,以北京轨道交通13号线为例,验证模型的有效性。第四,基于职住平衡理论,从居住空间分布优化角度建立拥挤路网下居民住宅分布的多目标优化模型,给出基于多目标值排序组合选择的遗传算法,并给出算例分析。第五,用数据包络分析方法(Data Envelopment Analysis, DEA),结合模糊数学,对居住就业空间分布与通勤交通的影响关系进行分析,建立中观评价模型。从就业空间分布优化角度,对非DEA有效的决策单元进行投影分析,计算改进的目标值。目标值的改进方向与居住就业平衡理论一致。最后,以北京市各行政区为例进行了评价分析。本论文的主要创新性成果如下：1、从微观角度研究居住就业与城市通勤交通出行关系,基于随机效用最大化理论,构建了居住和出行方式联合选择的网络广义极值模型,进行直接弹性和间接弹性分析,尝试刻画日益增长的交通拥堵情况的影响变化及其在不同的就业地模式下对居住再选址和出行方式转变的潜在影响；同时与小波神经网络模型的预测能力和解释能力加以对比。2、建立了基于最小二乘支持向量机回归的轨道交通沿线房地产价格预测模型,该模型能够在先验数据有限的情况下,较为准确地预测出城市轨道沿线房地产价格的分布规律。3、建立了轨道站点周边居住和出行方式联合选择模型,针对空间相关性问题,采用交叉分层Logit模型进行参数估计,定量研究轨道交通对居住选择和出行方式改变的潜在影响,有利于把握轨道站点周边居民出行空间分布特征,同时为合理规划轨道站点周边土地利用布局提供参考。4、构建了网络均衡状态下居住空间分布多目标优化模型,以居民居住地到工作地的平均出行距离成本最小,系统总出行时间成本最低,居民居住选址效用和社会效益最大作为目标函数,同时把路网平衡下的条件转化为优化目标的约束条件；在分析多目标优化Pareto最优解特点的基础上,提出了基于多目标值排序组合选择的多目标遗传算法对拥挤路网下居住空间分布进行优化,同时保证路网居住和就业的平衡关系,以此减少过量通勤交通,缓解交通拥堵,为交通规划和政策的研究提供参考。5、运用DEA方法对居住就业空间分布对交通出行的影响进行分析,建立中观评价模型。研究居住就业空间匹配均衡对通勤交通出行影响的程度,通过评价结果分析,能够识别出影响二者协调发展的关键因素,找出导致决策单元无效的原因,通过对非DEA有效的决策单元的投影分析对居住和就业空间分布进行优化。更多还原

【Abstract】 There is an even stronger correlationship between urban commuting and jobs-housing spatial distribution during the process of employment suburbanization and residential sprawl. Experience of both domestic and overseas witness the importance of understanding the interaction between transport and activity location for traffic congestion relief and for making practical and plausible transport policy as well. This dissertation aims to present approaches to represent the interaction relationship between urban commuting and residential-employment spatial distribution from microscopic and macroscopic perspective and propose programming models to optimize jobs-housing spatial distribution. Thus, providing some reference for urban transportation planning and urban spatial structure optimization.Firstly, previous studies on the relationship between urban commuting and jobs-housing spatial structure which provide useful theoretical foundation and inspiration for this dissertation are reviewed. Secondly, the interaction mechanism between urban commuting and jobs-housing spatial structure has been explored from both macroscopic and microcosmic perspective. NetworkGEV and WNN are respectively employed to model joint choice of residential workplace location and commute mode split. The predictive and representative capacity of these two models are compared. Thirdly, the impact of the construction of rail transit on residential location and commute mode choice is described by cross-nested loit simultaneous estimation model; meanwhile, LS-SVR is used to forecast real estate prices along urban rail transit in the light of small data samples. Rail transit line 13 in Beijing is taken as examples to illustrate the application of this model. Fourthly, based on jobs-housing balance theory, a multi-optimization model for residential location is established and solved by multi-objective genetic algorithm. Lastly, the relationship between urban commute and jobs-housing spatial structure is evaluated based on Data Envelopment Analysis (DEA) from mesoscopic perspective. A series of evaluating indexes are introduced. From empirical study of 18 Beijing districts, key influence factors of the relationship between urban commute and residence and employment configuration are identified and the non-effective decision making units are improved by projection analysis which is consistent with jobs-housing balance theory.The following are the main innovations of this dissertation: 1. A joint residential location and travel mode choice models under different employment destination scenarios are developed which representing the relationship between jobs-housing spatial distribution and travel to work from a microscopic perspective. Based on random utility maximization theory, discrete choice model specified as Network Generalised Extreme Value(NetworkGEV) has been employed to investigate the joint decisions of where to live and how to get to workplace which attempts to describe the change of aggravated traffic congestion and discover the potential change caused by residential relocation and travel mode shift under different employment location patterns. In addition, a WNN joint choice model is compared with NGEV model in terms of predictability and transferability.2. A LS-SVR model is developed to forecast real estate prices along urban rail transit in the light of the advantage of SVR dealing with small data samples. The kernel function is GaussianRBF providing a higher predictive accuracy than traditional hedonic price approach.3. On account of the advantages of cross-nested logit model structure to allow capturing the potential spatial correlation between spatially contiguous alternatives, cross-nested logit model is specified to estimate the joint choice of residential location and commute mode around the metro stations which help for the understanding the characteristics of spatial traffic distribution. The impact of the scenarios of travel time increasing on the choice probability switching is simulated in this model.4. A multi-objective optimization model is designed to optimize housing units in term of residential location under road network equilibrium. Multi-objective genetic algorithm emerges as the requirement to solve such kind of model. First, multi-objective functions composed of minimum average travel cost from home to workplace, minimum total commute cost and maximum utility of residential location are listed under the environment of road network equilibrium which is considered as constraints for multi-objective optimization model as well. Second, based on the characteristics analysis of multi-objective pareto optimum alternatives, the selection of multi-objective value ordering and combining is proposed to solve this multi-objective optimization model in which jobs-housing balance is remained, thus decreasing the excess commuting and relieving traffic congestion. Finally, a numerical example of residential location optimization is presented and the road network is configured to link the residential zones to workplace zones. This model is simulated by multi-objective genetic algorithm and is provided with a set of pareto optimum alternatives. 5. Data envelopment analysis is explored to establish the impact evaluation model and on the basis of fuzzy mathematics theory, the impact of residential and employment spatial configuration on the commute travel is correspondingly analyzed. Projection analysis is consequently carried out to improve non-effective DMUs and thus optimizing the employment spatial distribution.更多还原