城市是一个复杂的人地关系系统，城市可持续发展的核心是实现人地关系的协调。人类是城市的主体，城市中的餐饮、医疗等生活服务设施能够为人类的生存和发展提供不同类别的服务。因此，城市生活服务设施与居民之间的关系是城市人地关系系统的重要组成部分。生活服务设施可达性研究能够解析设施布局与居民分布之间的匹配关系，有助于指导设施布局优化，对城市的可持续发展至关重要。然而，传统的利用统计数据在区县尺度上研究设施可达性的方法，无法精确揭示城市内部小尺度区域的可达性差异。以电子地图兴趣点（Point of Interest, POI）数据为代表的大数据的兴起，为在街道甚至更小尺度上开展可达性研究提供了数据保障。此外，大数据大量、高频、多样化的特点拓宽了生活服务设施的分析视角，推动了城市生活服务设施可达性研究方法的革新。 论文基于可持续发展理念，以协调人地关系为目标，围绕“可持续发展—人地关系—可达性”的逻辑链条，以生活服务设施可达性为核心，深入挖掘大数据支持下可达性研究方法的改进，探索考虑不同类型生活服务设施特点的可达性评价方向：（1）通过剖析口碑对服务设施可达性的影响，提出考虑口碑影响的可达性评价模型；（2）针对服务设施的层次性结构，提出层次性可达性评价模型；（3）借助空间滤波模型，提出能够消除空间自相关影响的可达性影响分析模型。同时，以北京市为案例，应用本文提出的可达性评价模型，针对典型的经营性生活服务设施和公益性生活服务设施可达性进行评价。 首先，以餐饮设施为例，结合互联网上的网络口碑数据，运用考虑口碑影响的改进模型研究区域上餐饮设施可达性，对比经典模型评价结果，探究口碑对可达性的影响。论文揭示出北京市餐饮服务资源配置不均，设施在中心城区高度聚集的问题，并通过与经典模型评价结果的对比，发现考虑口碑影响时北京市餐饮资源不均衡分布的问题更加凸显，经典模型与考虑口碑影响的模型在评价结果上存在不同程度的偏差。 其次，以医疗设施为例，结合医疗设施POI数据，运用层次性可达性评价模型研究不同转诊模式下医疗设施可达性的变化；北京市医疗服务资源同样存在过度集中在中心城区的问题。不同转诊模式下层次性设施可达性评价模型结果表明：北京市目前可以在控制转诊率的前提下适度发展单向转诊模式，但不适宜发展双向转诊；仅凭转诊无法改变医疗资源在中心城区高度聚集的现状。

The city is a complex system of human-earth relationship. The city’s sustainable development relies on the coordination of human-earth relationship. Humans make up the main body of the city. Urban consumer (living) service facilities such as restaurants or hospitals are designated to provides services for human’s survival and development. The relationship between human and these facilities is a fundamental content of urban human-earth relationship system. Studies on the accessibilities of urban service facilities can reveal the relationship between facilities’ locations and the distribution of urban citizens, guide the location allocation of facilities and promoting the city’s sustainability. Previous studies often use statistical survey data to calculate facility’s accessibility on county or city scale. Studies on such large scale could not reflect the detailed situation within complex urban environment. Rising big data including point of interest data on digital map (POIs), mobile phone data etc. provide the data support for facility accessibility studies on finer scale. In addition, also stimulated the models of accessibility model. Based on the concept of sustainable development, aiming to the optimization of human earth relationship, following the logical chain of “sustainable development-human-earth relationship-accessibility”, the paper intensively explored further improvements of accessibility measures with the support of big data: (1) base on the influence of facility ratings, proposed rating based cumulative opportunity model; (2) regarding the hierarchical structure of facilities, proposed hierarchical three step floating catchment area model; (3) integrated eigenvector filtering model into hedonic regression to improve the accuracy of regression estimation. Meanwhile, the paper took Beijing as an example, chose representative private and public service facilities and conducted several case studies with proposed models mentioned above. Firstly, with the support of online rating, we used the rating based cumulative opportunity model to evaluate the accessibility of catering facilities of the study area and analyzed online ratings’ influence on facility accessibilities. The paper revealed that Beijing’s catering facilities were disproportional distributed and overly clustered in the central city. Though the comparisonof accessibility evaluation results between rating-based model and classic model, the paper found that the disproportional distribution problem of catering facilities grew server when facilities’ ratings were considered. Wihout considering the influences of facility ratings, the classic model would overestimate or underestimated catering facility’s accessibility Secondly, based on the POIs of medical facilities, we evaluated the accessibilities of medical facilities of the study area and investigated the difference of medical accessibilities under varies referral scenarios. The paper also revealed that the disproportional distribution problem also existed in Beijing’s medical facilities. Based on the results of hierarchical medical facility accessibility model, the paper concludes that: the study area could develope one-way referral scheme with the referral rate properly controlled and two-way referral scheme is not suitable for the study area, given the current allocation of medical facilities. Finally, the ESF-hedonic model is used to estimate the influences of multiple facilities’ accessibilities on housing prices. The paper found that eigenvector spatial filtering model could substantially improve the estimation accuracy of spatial regression. With eigenvector filtering model’s help, the regression result indicated that while accessibilities of both private and public facilities have significant influences on housing prices, the influences public facilities were generally greater.