城市的产生和发展是人类文明的重要体现。城市是人类各种社会经济行为的载体， 它为人们提供了安全保障，促进了知识与技能的交流与分享，进而促进了社会进步；同 时，人类的各种社会经济活动，也推动着城市的进一步发展，催生更大的聚集规模、更 高效的交通以及其他服务设施，以及适应新的技术发展的城市结构与功能。鉴于城市对 于人类社会的重要性，它早以成为各领域学术工作者的研究对象，大家试图从社会、经 济、环境、生态、建筑、规划等各个角度研究城市的发展，提出解决城市问题的可能方 案。然而过去的大部分研究通常只关注于城市的某一侧面，没有从关联与交互的视角研 究城市结构与功能的交互演化。 从系统科学的研究视角来考察，城市是一个典型的演化的复杂系统。人群的社会经 济行为与城市的方方面面紧密耦合、共同演化，造就了城市中各种结构与功能的涌现。 实际上，伴随着复杂性科学的兴起，相关概念逐步应用到定量化城市科学研究中，并产 生了许多有价值的成果，如分形城市理论、城市规模的齐普夫率、标度率、空间生长模 型等等。但总体来说，我们对于城市的描述与认知在很多方面仍然是不足的，缺乏能够 同时解释不同尺度下的复杂城市现象并进行有效预测的统一理论框架。随着科技的发展 与大数据时代的到来，我们得以获取到越来越丰富详细的城市数据，这为我们更加清晰 地把握城市脉搏、洞察纷繁城市现象背后的机制提供了机遇和挑战。本文融合大数据挖 掘技术以及关联和交互的系统科学研究方法，在基于大数据的定量化实证研究揭示唯象 规律、挖掘城市复杂系统的演化机制等方面开展研究，获得了以下研究成果。 1、针对与城市结构和功能密切相关的人群出行行为，本文首先进一步完善与发展 了手机大数据挖掘算法，通过细致的噪声过滤、时序与空间聚类、地点探测与标识等算 法，使得从海量数据中获取更细致精确的人口空间分布、用户出行行为、出发地-目的地 矩阵、交互活动信息成为可能。由于城市人群的出行行为与城市交通网络是实现城市社 会经济功能的重要基础，通过综合考虑人类行为以及道路特征，建立了更加合理的交通 效率评估的系统性指标—人口权重效率。已有的城市交通效率研究，要么只关注路网的 结构特征，包括拓扑特性和空间特性，如介数、最短路径、基于直线距离与导航距离之 比的迂回度等，要么只关注城市宏观尺度上人群的需求，如考虑目前通勤需求量与所谓 I 最优化需求之间差异的过剩通勤指标，没有综合考虑到城市中人群的动态特性以及底层 网络的特征。城市不只是一个静态的地点，它本身更是包含了动态的过程，居民何时、 何地、如何使用城市的基础设施网络将会极大地影响其效率。通过将这两种因素有机结 合，提出的人口权重效率这一概念既可以在城市级别对其整体交通效率进行评估，还可 用于城市内部及城际间低效道路的探测。最后还提出了一个简单的模型来解释观察到的 城市交通效率背后可能的机制。 2、提出了动态活跃人口概念，通过实证分析刻画了城市人口、路网、以及经济产出 的空间分布特征，进而建立了城市演化的动态空间生长模型，再现了普适的城市空间分 布的标度特征以及与城市规模相关的宏观标度关系。城市是人类各种社会经济行为的载 体，从复杂系统的角度看，人群的移动和交互行为是形成城市宏观结构与功能性质的基 础。(1) 为了更好地描述与估计城市中人际之间的交互，本文提出了动态活跃人口这一概 念，它综合考量城市区域的居住人口和工作人口，为进一步正确揭示人口分布与城市各 种社会经济行为的关系打下了基础；(2) 通过对城市大数据的细致分析，给出了城市人 口空间分布以及城市经济产出空间分布的函数关系，为建立城市演化机制模型提供了保 障；(3) 以城市演化进程中的人群聚集偏好、匹配效应为基本假设，进一步假定城市的 经济产出与人群及其交互密切相关，建立了城市演化的机制模型。与已有研究通常只分 别关注城市的形态、道路网络或者城市内部的经济活动不同，我们的模型关注城市内不 同子系统互相之间的关联以及影响。我们的模型仅仅基于抽象出的四条简单规则，就可 以重现出许多观察到的城市多尺度空间结构特征，诸如包括人口、道路、社会经济交互 在内的各种城市要素的空间分布与空间标度率、以及跨城市的宏观标度率，并且能够在 较高的精度上对于社会经济活动进行估计。同时我们的理论解析方法也是基于动态增长 的，而非过去城市研究中使用的全局平均场假设。模型还具有很强的扩展性，还可以用 来预测和分析诸如房价分布之类等其他受到广泛关注的城市相关的社会经济问题。 3、基于对人的移动与交互行为的研究结果，探讨了其对于复杂城市系统上传播动力 学的影响。已有的复合种群模型大都基于对城市节点内部进行均一化的平均场假设，我 们发现每一个城市中人口的活跃程度、接触强度以及移动模式都是大不相同的，而这样 的接触和移动异质性对于流行病传播过程有着重要影响。因此，我们将人类移动以及人 群的交互这两个主要因素纳入了以城市为节点构建的非线性复合种群模型，进而证明由 于人类移动及活跃强度的不同，基本生成数大于1对于流行病的爆发并非必要条件。根 据基于人类移动流量定义的等效距离，我们还能将实际地理空间上复杂的传播过程还原 为高维流型上相对简单的扩散过程，进一步展现了人类移动和交互行为对传播过程的影 响。 本文的研究方法与结果加深了我们对于城市演化及其结构与功能的理解，为城市研 II 究以及其他领域的复杂系统的研究提供了新的视角，具有重要的理论意义；同时，对城 市演化机制的认识，对于城市规划、交通路网设计等也具有重要的实践价值，有着广泛 的应用前景。

The emergence and the development of the city is a significant advance in human culture. Cities have become the carrier of various human socio-economic activities by providing effective security assurance and promoting the communication and sharing of knowledge and skill which all further promotes the development of the society. Meanwhile, all kinds of human socioeconomic activities push the further development of the city, resulting in larger size of agglomeration, more efficient transportation, other service facilities as well as new structure and function of the city which adapts to the development of state-of-the-art technologies. Given the importance of the city to human society, it has attracted great attentions from scientists in various academic fields (such as sociologists, economists, environmentalists, ecologists, architects, urbanologists), who had tried to propose theories to urban development and possible solutions to urban issues from their own theoretical perspectives. However, most of the previous researches only focused on a certain aspect of the city, yet didn’t give enough attention on the co-evolution of the urban structure and function. From a systems science perspective, the city is a typical evolving complex system. The socioeconomic interactions are strongly cohesive with co-evolution and various aspects of the city which in turn leads to the emergence of abundant structures and functions of the city. In fact, along with the rise of the complexity science, related concepts have been gradually applied to quantitative studies of urban science, producing a great many of valuable theories and findings, such as fractal city, Zipf’s law on the size of global cities, scaling laws, spatial growth model, etc. Yet we are still in lack of a more comprehensive understanding of many aspects of the city and a unified theoretical analytical framework for explaining and predicting complex urban phenomena at various spatial scale. In the wake of the fast development of technology and dawning of the exciting big data era, we are able to accumulate much richer and more abundant multisources urban data than any past times, which enable us to precisely grasp the pulse of the city and gain insights of mechanisms behind all kinds of complex urban phenomena. It is both a great opportunity and challenge. This dissertation is fusing the bid data mining algorithms and systems science perspective which focuses on relation and interaction to reveal the phenomenological laws through quantitative empirical studies and mining the evolutionary mechanisms behind complex urban systems. It leads us to the following findings. (i) Dedicated to the human travel behavior which is closely related to the structure and function of the city, this dissertation first lands on developing and improving big cellphone data mining algorithms. Through careful noise filtering and temporal, spatial clustering algorithms, it becomes possible to accurately reveal the information from massive cellphone data of where and when the user was and how long he/she stayed in a specific location. Then we can aggregate all their trips to get the Origin-Destination (OD) matrix of the city as well as get a better estimation of the active population distribution. Due to the fact that human travel behavior and urban transportation network are fundamental to achieve urban socioeconomic functions, by fusing these two factors we establish a more reasonable and systematical evaluation indicator – Population-Weighted Efficiency (PWE). Previous researches on transportation efficiency either only focus on the structural and spatial features of the underlying network (such as betweenness, inverse of shortest distance, route factor that is the mean ratio of the distance along edges and Euclidean straight line, etc.) or concentrate on the travel demand at macro urban scale (such as job-housing balance or excess commuting [i.e., the difference between the average actual commuting time (or distance) and the theoretical minimum average commuting time (or distance)]). However, the first kind of attempts has not considered the real population distribution, while the second suffers from the use of the statistically improbable variable (minimum average commuting distance) as a benchmark and neglecting the specific features of the underlying networks. The city is not just a static place but also a dynamic process, when, where and how the citizens use the infrastructural networks strongly influence its efficiency. By incorporating both human travel and the network factors, the proposed new concept of PWE is able to investigate the efficiency in a more systematical way. PWE is a better indicator than excess commuting on evaluating the average commuting time and give a more accurate evaluation of the transportation efficiency of the city, and it can be applied to detect inefficient local routes which were the weakness of excess commuting index. And apart from the cellphone data, our analysis framework can be carried out by open source data. In the end, we also propose a simple null model to explain the possible mechanisms behind these empirical findings on urban transportation efficiency. (ii) This dissertation proposes a brand new concept of dynamic active population (AP), and through empirical studies reveals the spatial distribution characteristics of population, road networks and socioeconomic output. And then further builds a dynamic spatial urban growth model which can reproduce the general spatial scaling laws within cities and scaling laws on macro variables across cities. Cities are the carriers of all kinds of socioeconomic activities, and from a systems science perspective, the human interactions are the very nature and driving force of emergent urban structures and functions. (1) In order to give a better description and estimation of human interactions within cities, the proposed concept of AP , which is composed by both residential and working population according to their active duration in a certain location, paves the way for further correctly revealing the population and various socioeconomic activities distribution; (2) Through careful analysis of big urban data, we give the formula of such distributions and provide solid evidence for further building the urban evolutionary model; (3) Based on preferential spatial attraction and matching growth mechanisms which are abstracted from real urban evolution process and empirical studies, and together with road network generation rule as well as socioeconomic output rule, we build the complete spatial urban evolutionary model. Different from previous works which tend to just focus on one aspect of the city (such as the morphology, street network or economic activities), our model focus on the impacts of relation and interactions between different urban elements. The proposed model is only based on those 4 simple abstract rules, yet it can explain not only spatial scaling behaviors of urban elements within the city, but also the origin of the scaling laws on macro variables across cities, as well as give predictions on socioeconomic activities at a high spatial resolution. Besides, our theoretical analysis method is original which is based on growth instead of static mean-field assumptions. And due to the simplicity and high extendability of our model, in the framework, we can also investigate other socioeconomic problems that attracted great attention, such as the spatial distribution of house price, etc. (iii) Based on the results of human travel and interaction, this dissertation studies the impacts of these two factors on epidemic spreading on complex urban systems. Previous meta-population models all assume a well-mixed situation within the city, yet we discover that the interaction activity intensity, human mobility behaviors in cities are quite different from each other, and such heterogeneities are strongly influencing the epidemic spreading process. So we take human mobility and human interactions these two main factors into the nonlinear meta-population model whose nodes are cities. We prove that due to such effects, the requirement of the basic reproductive number larger than 1 for triggering the epidemic is neither sufficient nor necessary. And according to the proper distance defined by human mobility, we can map the observed complex spreading process on real geographical space roughly back to a simple diffusion process on a higher dimensional manifold which further manifests the great impacts of human mobility and interactions on epidemic spreading dynamics. The methodology and findings of this dissertation are of great theoretical value which deepens our understanding of urban structure and function as well as their co-evolution and provides new insight and perspective to urban studies and other related complex systems researches. Meanwhile, the results also have a broad range of application which can benefit urban planning, traffic engineering, transportation network design, infectious disease spreading and other fields.