Extracting Traffic Events and Human Mobility Patterns in Geosocial Media Data for Assessing Real-time Road Traffic

大城市中心和新兴城市区域，以及世界其他地区，在继续快速增长的同时，面临着一系列复杂的挑战，如交通拥堵。世界上许多最大的城市，包括加拿大的温哥华和多伦多，都是交通拥堵最严重的城市之一。随着日益严重的交通拥堵导致延误和驾车者的健康问题，并产生额外的空气污染，可靠的交通信息对世界各地的旅行者来说是关键。这项拟议的研究将开发新的交通事件检测方法，以便使用海量社交媒体数据提供此类实时信息，并将构建额外的工具，将人类移动模式整合到短期交通拥堵预测中，这将有助于预测出行者未来15-60分钟的道路交通状况。本研究的成果(模型、方法和工具)对新一代城市规划者、城市交通管理者和研究人员具有很大的参考价值。 拟议研究的短期目标将集中在三个关键研究部分：1)在不同的空间和时间尺度上模拟人类活动模式，以研究它们与经常性和非经常性交通情况的互动；2)检查和评估不同的交通和交通相关事件检测方法，以开发基于时空语义的检测方法；以及3)评估和探索社交媒体数据以及相关模型、工具和工作流如何通过开发面向服务的协作式开放平台来支持实时道路交通信息的提供。该平台将使研究人员、交通管理人员和决策者能够直接和及时地获取实时交通数据，这些数据主要提取自各种社交网络、交通信息服务，以及用于交通相关研究的合适的计算和分析模型和工具。 我的研究计划的长期目标是开发一个实时道路信息基础设施，该基础设施能够可视化、发布和向目标用户推送近实时的交通信息，并提供最优路径服务，并且能够预测考虑非经常性交通情况和考虑周期性模式的长期交通拥堵。这样的基础设施可以作为一个基于Web的系统来实施，该系统集成了云计算、Web服务和地理信息系统，用于交通管理决策和交通规划。 拟议的试验性开放平台，将模型、方法和工具的实施作为服务，将建设强大的道路交通信息和服务共享能力，并将为研究共享、HQP培训以及与加拿大政府机构和行业的合作开辟新的机会。

Large urban centers and emerging city regions, as well as others in the world, face a series of complex challenges such as traffic congestion as they continue to grow at a rapid pace. Many largest cities around the world, including Canadian cities such as Vancouver and Toronto, are known among top worst for traffic congestion. With increasing traffic congestion causing delays and health problems for motorists and producing extra air pollution, reliable traffic information is key for travelers all over the world. This proposed research will develop new traffic event detection methods for providing such real-time information using massive social media data, and also will build extra tools for integrating human mobility patterns into short-term traffic congestion prediction that will help predict traveler's road traffic conditions in next 15-60 minutes ahead. The results (models, methods and tools) of this research can be of great use by a new generation of urban planners, city traffic managers and researchers. The short-term objectives of the proposed research will focus on three key research components that will: 1) model human activity patterns at various spatial and temporal scales to study their interactions with recurrent and non-recurrent traffic situations; 2) examine and evaluate different traffic and traffic-related event detection methods to develop a space-time-semantics based detection approach; and 3) assess and explore how social media data together with related models, tools and workflows can support the provision of real-time road traffic information by developing an collaborative service-oriented open platform. The platform will give researchers, traffic managers and decision makers direct and timely access to real-time traffic data, primarily extracted from various social networks, traffic information services, and suitable computational and analytical models and tools for traffic-related studies. The long-term goal of my Research Program is to develop a real-time road information infrastructure that visualize, publish and push near real time traffic information to target users and provide optimal routing services, but also capable of predicting real-time traffic congestion taking into non-recurrent traffic situations and long-term traffic considering recurrent patterns. Such an infrastructure can be implemented as a web-based system that integrates cloud computing, web services and GIS for traffic management decision making and transportation planning. The proposed experimental open platform, with implementation of the models, methods and tools as services, will build strong capacity for sharing road traffic information and services and will open up new opportunities for research sharing, HQP training and collaborations with Canadian government agencies and industry.