Mining, Fusion and Modeling of Truck Big Data for the development of Agent-Based Microsimulation Models

现代供应链流程在很大程度上依赖于不同制造商、分销商和零售商之间的货物转移。到目前为止，卡车是加拿大主要的货物运输方式。近年来货运活动的增加突出了了解卡车货运活动的重要性。此前研究货运量的努力一直依赖于调查数据。然而，这些调查通常对分析师来说很昂贵，对受访者来说也很耗时。过去几年来，一种新兴的调查替代方法是分析全球定位系统(GPS)数据。20多年来，运营商一直使用GPS应答器来实时识别卡车的位置，并确定货物到达的时间。虽然GPS数据为研究人员提供了极好的机会，但它们在交通建模中的使用具有挑战性，原因有两个：1)这些GPS记录最初不是作为交通模型和分析的输入，2)获得的GPS数据的数量、速度和多样性转化为数十亿条记录(即大数据)，必须进行分析。因此，需要新的方法和技术从被动GPS记录中挖掘大数据，并将这些数据与其他来源融合，以便了解和模拟卡车移动和空间供应链过程。 为此，该研究计划的具体目标如下：1)使用描述7个月(2015年9月至2016年3月)期间(2015年9月至2016年3月)60,000辆加拿大卡车移动的大数据，促进加拿大货运和供应链流程的当前知识状况；2)开发、实施和测试新的数据挖掘方法，可用于提取有关卡车移动和贸易流动的有意义的模式和知识；3)开发一个新的软件平台，嵌入目标2中设计的方法，以可视化和表征加拿大境内以及加拿大和美国之间的卡车移动的性质；4)模拟货运流动，以了解托运人的出行行为，以及影响货物通过运输设施流动的潜在因素；5)开发货运、贸易流动和供应链流动的综合模拟和微观模拟预测模型，以研究可以改善主要拥堵走廊和跨境点交通表现的政策。 拟议研究计划的新颖性有两个：1)这是加拿大第一个分析描述2015年和2016年加拿大大量卡车移动的大数据的研究计划；2)据我们所知，这是加拿大唯一一个寻求开发和应用一套新的数据挖掘和融合方法的研究计划，这些方法来自GPS的大数据，并使用结果设计货运微观模拟模型。

Modern supply chain processes rely heavily on goods transferred between different manufacturers, distributors and retailers. Trucks are by far the dominant mode of transportation for moving goods in Canada. Increased freight transportation activities in recent years underline the importance for understanding freight movements by trucks. Previous efforts to study freight movements have relied on data coming from surveys. However, these surveys are usually expensive for the analyst and time consuming for the respondent. An emerging alternative to surveys over the past few years has been the analysis of Global Positioning Systems (GPS) Data. Carriers have been using GPS transponders for over 2 decades to identify the location of their trucks in real-time and to determine when their goods get delivered. While GPS data provide remarkable opportunities to researchers, their usage in transportation modeling is challenging for two reasons: 1) these GPS records were not originally intended as an input for transportation models and analysis, and 2) the volume, velocity and variety of the obtained GPS data translates into billions of records (i.e. big-data) that have to be analyzed. Therefore, a need exists for novel methods and techniques to mine big-data from passive GPS records and fuse these data with other sources in order to understand and model truck movements and supply chain processes over space. To this end, the specific objectives of this research program are as follows: 1) Advance the current state of knowledge on freight movement and supply chain processes in Canada using a big-data that depict the movement of 60,000 Canadian trucks over a period of 7 months (September 2015 - March 2016); 2) Develop, implement and test new data mining methods that can be used to extract meaningful patterns and knowledge on truck movement and trade flows; 3) Develop a new software platform that embeds the devised methods from objectives #2 to visualize and characterize the nature of truck movements in Canada and between Canada and the US; 4) Model freight transportation movement to understand the travel behavior of shippers, and the underlying factors affecting the movement of cargo through transportation facilities; 5) Develop integrated simulation and microsimulation predictive models for freight transportation, trade flow and supply chain movements, to examine policies that could improve the performance of traffic on major congested corridors and cross-border points. The novelty of the proposed research program is twofold: 1) this is the first research program in Canada to analyze a big-data that depicts the movement of a large number of Canadian trucks for the years 2015 and 2016; and 2) to our knowledge, this is the only research program in Canada that seeks to develop and apply a suite of new data mining and fusion methods on big-data coming from GPS and use the outcomes to devise freight microsimulation models.