Peeking inside the black-box: Explainable machine learning applied to household transportation energy consumption

Sustainability policies to mitigate transportation energy impacts on the urban environment are urgently needed. Energy prediction models provide critical information to decision-makers who develop sustainability policies to reduce energy use and emissions. We present a transportation energy model (TEM) that uses Explainable Artificial Intelligence (XAI) methods to predict household transportation energy consumption in this study. The TEM model uses data-driven approaches for household transportation energy prediction. Machine learning techniques in artificial intelligence (AI) predictive modeling have become popular due to their ability to capture nonlinear and complex relationships. On the other hand, developing comprehensive understanding the inference mechanisms in AI models and ensuring trust in their predictions is challenging. This is because AI models are mostly of high complexity and low interpretability; in other words, they are black-box models. This study presents a case study of how model transparency and explanation can be generated using the Local Interpretable Model-Agnostic Explanation (LIME) to support advanced machine learning techniques in the transportation energy field. The methodology has been implemented based on the Household Travel Survey (HTS) data, which is used to train the artificial neural network with a relatively high degree of accuracy. The importance and effect (local explanation) of HTS inputs (such as household travel, demographics, and neighborhood data) on transportation energy consumption for specific traffic analysis zones (TAZs) are analyzed. The results are valuable to promote intelligent and user-friendly transportation energy planning models in urban regions across the world.     

Modelling household travel energy consumption and CO2 emissions based on the spatial form of neighborhoods and streets: A case study of Jinan,China

This paper presents a direct modelling approach to evaluate household travel energy consumption and CO₂ emissions based on the spatial form of neighborhoods and streets. It integrates one multinomial logit model and four double hurdle models to predict travel outcomes: vehicle ownership portfolio choice, car travel distance, bus travel distance, motorcycle travel distance and ebike travel distance. The energy consumption and CO₂ emissions are then estimated using these travel outcomes. A full application of the modelling approach is demonstrated through a pilot project in Jinan, China. A holdout validation is also performed to address the over-fitting problem of models calibrated from the training set data. Results show that the proposed approach is operational and appropriate for sketch planning applications to promote clean energy and low carbon city planning in urbanizing China.     

基于Spark平台城市出租车乘客出行特征分析

从海量出租车GPS轨迹数据中挖掘和分析城市出租车乘客的出行特征,可以为城市交通管理者和出租车行业管理者在城市交通规划与管理、城市交通流均衡与车辆调度等方面提供决策依据.基于Spark大数据处理分析平台,选择YARN作为资源管理调度系统,采用HDFS分布式存储系统,对出租车GPS轨迹数据进行挖掘.给出了基于Spark平台的出租车乘客出行特征的挖掘方法,包括出租车乘客出行距离分布、出租车使用时间分布及出租车出行需求.实验结果表明,基于Spark平台分析方法能够快速且准确的分析出出租车乘客出行特征.     

A GPS/GIS method for travel mode detection in New York City

Handheld GPS provides a new technology to trace people’s daily travels and has been increasingly used for household travel surveys in major cities worldwide. However, methodologies have not been developed to successfully manage the enormous amount of data generated by GPS, especially in a complex urban environment such as New York City where urban canyon effects are significant and transportation networks are complicated. We develop a GIS algorithm that automatically processes the data from GPS-based travel surveys and detects five travel modes (walk, car, bus, subway, and commuter rail) from a multimodal transportation network in New York City. The mode detection results from the GIS algorithm are checked against the travel diaries from two small handheld GPS surveys. The combined success rate is a promising 82.6% (78.9% for one survey and 86.0% for another). Challenges we encountered in the mode detection process, ways we developed to meet these challenges, as well as possible future improvement to the GPS/GIS method are discussed in the paper, in order to provide a much-needed methodology to process GPS-based travel data for other cities.     

基于MapReduce的城市道路旅行时间实测计算

城市道路旅行时间计算一直是智能交通系统中研究的核心问题之一,准确高效的旅行时间计算可以有效地帮助道路管控,减少交通拥挤.然而面对巨大而且快速增长的城市道路交通检测数据,如何将分布式计算模式融合到传统的旅行时间计算问题中已成为一个亟待解决的问题.论文基于海量道路车牌识别数据,设计了基于MapReduce编程模型的城市道路旅行时间实测计算的算法.并利用Hadoop环境进行了实现,可以支持对自定义路段集下不同时间段道路旅行时间的计算.通过实验证明,相对于传统的旅行时间计算方式,在计算时间上基于MapReduce的旅行时间计算模式可以提高十倍以上.     

双参卷积理论模型预测交通通行时间

随着城市经济的发展和人们生活节奏的加快,智慧交通领域针对出行时间的研究已经成为热点问题。出行前预估行程中的通行时间便于人们更合理地规划出行路径,基于时间状态特征的路径规划就是解决交通问题的重要手段之一。现有模型多关注于车辆到达时间或多结合于真实历史时间数据进行预测,对浮动车的运行状态、车速等是否对时间存在影响的问题研究较少。基于此现状,提出了一种基于状态特征的道路时间预测模型,在固定时段内,利用出租车载客与否情况对轨迹数据进行深度相关性分析,结合车辆行驶速度构建一个基于密度划分的双参卷积理论模型,用得到的最终速度值对通行时间进行计算。实验结果表明该模型算法与传统时间预测算法相比有更高的精确度和实用性,提高了人们对出行安排的合理化和层次化,对制定城市道路出行策略具有重要的意义。     

基于公共交通智能卡数据的可视化分析

针对现有公共交通数据的可视分析方法很难在不同空间粒度下对乘客时空分布、客流时空分布、区域间客流时序变化进行多任务分析的问题,设计实现了一个多视图融合的可视化分析系统。该系统结合城市公共交通的智能卡数据、车辆GPS数据、地铁和公交线路信息,利用出行链路模型和基于出行时空特征的回归模型完成了乘客起讫点（origin-destination,OD）推断;然后,设计了层次聚类的地图可视化方法,结合了融合方位信息的玫瑰图和动态对比堆叠折线流图来分析各区域间的客流时序特点、关联关系;最后,利用真实的深圳市公共交通数据的可视分析结果验证了系统的有效性。     

Using simulated annealing to minimize fuel consumption for the time-dependent vehicle routing problem

The vehicle routing problem (VRP) has been addressed in many research papers. Only a few of them take time-dependent travel speeds into consideration. Moreover, most research related to the VRP aims to minimize total travel time or travel distance. In recent years, reducing carbon emissions has become an important issue. Therefore, fuel consumption is also an important index in the VRP. In this research a model is proposed for calculating total fuel consumption for the time-dependent vehicle routing problem (TDVRP) where speed and travel times are assumed to depend on the time of travel when planning vehicle routing. In the model, the fuel consumption not only takes loading weight into consideration but also satisfies the “non-passing” property, which is ignored in most TDVRP-related research papers. Then a simulated annealing (SA) algorithm is proposed for finding the vehicle routing with the lowest total fuel consumption. An experimental evaluation of the proposed method is performed. The results show that the proposed method provides a 24.61% improvement in fuel consumption over the method based on minimizing transportation time and a 22.69% improvement over the method based on minimizing transportation distances.     

数据仓库在城市公共建筑能耗管理中的应用

当前众多城市公共建筑能耗监测系统中已收集了大量的建筑能耗数据。针对这些数据源存在的各自独立而且分散,不能够提供全局的数据分析环境,不能够有效支持建筑能耗的评估与建筑节能的研究等问题,提出将数据仓库技术应用于城市公共建筑能耗监管系统的解决方法。通过对建筑能耗监测系统的研究,以及对建筑能耗管理的应用需求的调研,建立城市级公共建筑能耗数据仓库的多维数据模型,对主题设计、指标设计和维度模型设计进行了探讨,并在实验阶段已成功构建了某高校公共建筑能耗数据仓库的实例。实验结果表明,该方法能够有效地为建筑能耗的管理与决策提供良好的数据分析环境。     

Network-based Accessibility Measures for Vulnerability Analysis of Degradable Transportation Networks

Accessibility is a fundamental concept in transportation analysis and urban planning. Typically, accessibility refers to the ‘ease’ of reaching opportunities for activities and services and can be used to assess the performance of a transportation and urban system. In this paper, we present network-based accessibility measures for assessing vulnerability of degradable transportation networks. The network-based accessibility measures consider the consequence of one or more link failures in terms of network travel time or generalized travel cost increase as well as the behavioral responses of users due to the failure in the network. To model different dimensions of travel behavioral responses, a combined travel demand model is adopted to estimate the long-term equilibrium network condition due to network disruptions. Numerical examples are conducted to demonstrate the feasibility of the proposed vulnerability measures for assessing degradable transportation networks. The results indicate that the accessibility measures derived from the combined travel demand model are capable of measuring the consequences of both demand and supply changes in the network and have the flexibility to reflect the effects of different travel choice dimensions on the network vulnerability.