Informs Annual Meeting 2017

WB66

INFORMS Houston – 2017

WB64

2 - Investigating the Impacts of Information Sharing on Intermodal Infrastructure Investment Decision-making Process under Demand Uncertainty Irina Benedyk, Purdue, 2411 Neil Armstrong Drrive, # 2A, West Lafayette, IN, 47906, United States, birina@purdue.edu, Srinivas Peeta We investigate the impacts of investment-related information shared between private sector (that is, timing, expansion and operational strategies) on intermodal infrastructure investment decision-making process under demand uncertainty. In addition, we study the impacts of public-private partnership arrangements (which include risk allocation, financial burden and asset ownership) on intermodal infrastructure investment behavior of private sector under market competition. The study findings will help to understand the value of information in intermodal infrastructure investment decision and its impact on investment related uncertainties that affect the return of the investment. 3 - Reliable Facility Location Design under Disruption Risks and Congestion in Continuous Spaces Zhaodong Wang, University of Illinois, 502 E Michigan Ave, Apt 21, Urbana, IL, 61801, United States, zwang137@illinois.edu, Siyang Xie, Yanfeng Ouyang This paper develops analytical methods to evaluate the expected system performance with considerations of facility disruptions under a class of continuous traffic equilibrium. Closed-form solution can be evaluated in polynomial time for heterogeneous independent disruptions. A Lagrangian- relaxation-based optimization framework is proposed based on the analytical results to optimize the facility location design. 4 - Best and Worst Values of the Optimal Cost of the Interval Transportation Problem Monica Gentili, University of Louisville, JB Speed Building, Room 304, Louisville, KY, 40292, United States, monica.gentili@louisville.edu, Raffaele Cerulli, Ciriaco D’Ambrosio We study the transportation problem with interval supply and demand parameters (ITP), that is, the transportation problem where supply and demand are uncertain and vary in given ranges. We are interested in determining the best and the worst values of the optimal cost of ITP among all the realizations of the uncertain parameters within the intervals. We prove some general properties of the best and the worst optimum values and propose efficient algorithms to determine them. 371A Traffic and Pedestrian Modeling Sponsored: Transportation Science & Logistics Sponsored Session Chair: Rui Ma, University of California, Davis, 1930 Sycamore Ln, #10, Davis, CA, 95616, United States, drma@ucdavis.edu 1 - A Comprehensive Assessment of Various Factors on Driving Ability of Individuals Maxim A. Dulebenets, Florida A&M.University-Florida State University, 2300 Bluff Oak Way, Apt. 8408, Tallahassee, FL, 32311, United States, mdlbnets@gmail.com, Olumide Abioye, Eren Erman Ozguven, Ren Moses, Walter Boot, Thobias Sando The majority of models used in transportation planning for estimating travel time and other performance indicators are primarily based on the traffic flow and roadway geometric characteristics. However, in reality there is a variety of other factors that have to be considered (e.g., driver characteristics, weather conditions, temporary attributes, etc.). This study performs a state-of-the-art review of the academic literature to identify a wide spectrum of different factors that may affect the driving ability of individuals. 2 - Investigating Drivers’ Perception of Travel Time Reliability using a Full-scale Driving Simulator Zohreh Rashidi Moghaddam, Morgan State University, 1700 E Cold Spring Ln, Baltimore, MD, 21251, United States, zoras1@morgan.edu, Mansoureh Jeihani This study investigates travelers’ perception of travel time reliability using a full- scale high-fidelity driving simulator. Participants in this study experienced driving in a realistic network in Maryland to make a route choice decision in different scenarios based on the travel time information shown on a Variable Message Sign. Driving experience was also accompanied by a stated preference (SP) survey with similar scenarios to determine their willingness to detour on a Likert scale. Multinomial Probit Model and Ordinal regression has been applied to model travelers route choice behavior and their willingness to detour respectively. WB66

370E Machine Learning Algorithms and Applications in Dynamic Environments Sponsored: Data Mining Sponsored Session Chair: Seoung Bum Kim, sbkim1@korea.ac.kr 1 - Dynamic Adjustment of Dispatching Rule Selection Sangmin Lee, Korea University, Seoul, Korea, Republic of, smlee5679@gmail.com, Seoung Bum Kim In the semiconductor industry, dispatching rule selection is a challenging task because of high uncertainty caused by a gap between planned capacity and actual capability. This study presents a dynamic dispatching rule selection method with multivariate multiple regression. To demonstrate the effectiveness and applicability of the proposed method, we conduct simulation study using real data from semiconductor fabrication plant in one of the leading semiconductor companies in South Korea. 2 - Label Propagation Algorithm with Consensus Matrix Jaehong Yu, Korea University, Seoul, Korea, Republic of, dreamer77445@gmail.com, Seoung Bum Kim This paper proposes a consensus matrix-based label propagation algorithm for semisupervised classification. The proposed algorithm constructs a consensus matrix to accommodate various characteristics of data and propagates label information using the iterative label fitting algorithm. The experimental results confirm that the proposed algorithm demonstrated superior performance compared to the competitors. 3 - Multi-co-training for Document Classification using Multi-views Donghwa Kim, Korea University, 145 Anam Roo, Seoul, 02841, Korea, Republic of, donghwa89@korea.ac.kr, Deokseong Seo, Suhyoun Cho, Pilsung Kang The purpose of document classification is to assign the most appropriate label to given documents. The main difficulties of document classification are insufficient label information and irreversible document representation. In this paper, we propose a multi-co-training based of different representations to improve the performance of document classification. In order to increase the difference of feature sets, we transform a document based on three different representation methods: term frequency-document frequency (TF-IDF) based on bag-of-words scheme, topic distribution based on latent Dirichlet allocation (LDA), and neural network-based document embedding known as Doc2Vec. 4 - Autoregressive Forests for Multivariate Time-series Modeling Kerem Tuncel, Boaziçi University, Istanbul, Turkey, keremsinantuncel@gmail.com, Mustafa Gokce Baydogan Multivariate Time Series (MTS) modeling has received significant attention in the last decade. E cient representations are required to deal with the high dimensionality of MTS. However, most of the traditional approaches assume independence or linear dependence between the variables of MTS which is unrealistic. To handle these problems, we propose an autoregressive tree based ensemble approach that can model the nonlinear behavior embedded in the time- series. An error-based representation based on the learned models is the basis of the proposed nonparametric vector autoregression approach. This is very similar to time series kernels used for multivariate time series classification problems.

WB65

370F Freight Transportation Planning and Design Sponsored: TSL, Freight Transportation & Logistics Sponsored Session

Chair: Monica Gentili, IPAT, Georgia Institute of Technology, 75 5th Street NW, Suite 600, Atlanta, GA, 30308, United States, mgentili@unisa.it 1 - An Online Cost-sharing Mechanism for Horizontal Supply Chains Maged M. Dessouky, University of Southern California, 3715 McClintock Avenue, GER.240, Los Angeles, CA, 90089-0193, United States, maged@usc.edu, Han Zou, John Gunnar Carlsson This research addresses the cost allocation problem in a real-time cost sharing transportation system, which results from horizontal cooperation among multiple suppliers. We formulate the cost allocation problem for the dynamic vehicle routing environment, where only part of the customers are known in advance, and the rest become known in real time. We propose an online cost-sharing mechanism coupled with specially designed cost functions. The mechanism can be incorporated with a look-ahead dynamic vehicle routing framework that explicitly forecasts future customer requests.

497