Identification of travelers' transportation modes is a fundamental step for various problems that arise in the domain of transportation such as travel demand analysis. In this study, we aim to identify users' transportation modes purely based on their GPS trajectories. A majority of studies have proposed mode detection models based on hand-crafted features using only labeled GPS trajectories. To address such shortcomings, we propose a deep SEmi-Supervised Convolutional Autoencoder (SECA) architecture that can not only automatically extract relevant features from GPS tracks but also exploit useful information in unlabeled data. The SECA integrates a convolutional-deconvolutional autoencoder and a convolutional neural network into a unified framework to concurrently perform supervised and unsupervised learning. The two components are simultaneously trained using both labeled and unlabeled GPS trajectories, which have already been converted into an efficient representation for the convolutional operation. An optimum schedule for varying the balancing parameters between reconstruction and classification errors are also implemented. Our experimental results demonstrate the superiority of the proposed model, the new representation for GPS trajectories, the hyperparameter schedule, and the model configuration over the state-of-the-art semi-supervised and supervised methods and alternative settings with respect to metrics such as accuracy and F-measure.

The resiliency of infrastructure, particularly as related to transportation networks, is essential to any society. This resiliency is especially vital in the aftermath of disasters. Recent events around the globe, including Hurricane Katrina and significant seismic events in Haiti, Chile, and Japan, have increased the awareness and the importance of resiliency. Transportation systems are key to response and recovery. These systems must withstand stress, maintain baseline service levels, and be stout enough in physical design and operational concept to provide restoration to the system. Analysis of a transportation network's resiliency before a disruptive event will help decision makers identify specific weaknesses within the network so that investments and improvement projects are prioritized appropriately. Previous research in quantification of network resiliency was expanded into a proposed methodology, through which understanding and applying concepts of network resiliency could preclude many devastating effects of destabilizing events and preserve the quality of life and economic stability.

Every day dependence on transportation grows as local, regional, national, and international independence increases. Resilient transportation systems are needed to secure the highest possible level of service during disruptive events, including natural disasters and those caused by humans. To prepare for these events, decision makers need guidance to determine what investments are likely to improve the resiliency of their networks, which are often hampered by limited resources. To date, such guidance has been primarily qualitative. This paper presents a methodology to quantify resiliency, under preevent conditions, by use of a fuzzy inference approach. This methodology expands on previous work by the authors and others, by refining the definitions of key variables, adjusting model interactions, and increasing transparency between metrics. The paper includes a case study in which the methodology is applied to a disruptive event in Santo Domingo, Dominican Republic. The case study illustrates the methodology's ability to (a) evaluate the extent to which the Dominican Republic's transportation network is prepared for a disruptive event, (b) help select investments that have the potential to increase the resiliency of the network, and (c) provide outputs that will support a variety of current economic analysis strategies, allow comparison of different investment scenarios, and facilitate decision making. The paper concludes with a sensitivity analysis that shows the effects of alternative investments on the network.