Informs Annual Meeting 2017

TD72

INFORMS Houston – 2017

TD72

TD73

372A Optimization, Nonlinear Programming Contributed Session Chair: Ragheb Rahmaniani, Université de Montréal, Montréal, QC, Canada, Ragheb.rahmaniani@gmail.com 1 - Team Orienteering Problem with Stochastic Profits and Risk Constraints Hadi Feyzollahi, SUNY University at Buffalo, Department of Industrial Eng., 316 Bell Hall, Buffalo, NY, 141260, United States, hadifeyz@buffalo.edu, Jose Luis Walteros We proposed a dynamic programming approach to tackle the Team Orienteering Problem over graphs with stochastic profits and risk constraints. We study both discrete and continuous probability distributions to model the profits and test different approaches to compute their convolutions. We further presented a case study to illustrate the results. 2 - The Multiperiod Network Design Problem Ioannis Fragkos, Rotterdam School of Management, Burg. Oudlaan 50, Rotterdam, 3062 PA, Netherlands, fragkos@rsm.nl, Jean-Francois Cordeau, Raf Jans We devise decomposition methods to solve large-scale instances of multi-period network expansion problems. For capacitated networks, we devise a custom heuristic procedure combined with arc-based Lagrange relaxation. For uncapacitated networks, we employ Benders decomposition, where the subproblems are decomposable per period and per commodity. We formulate the problem of generating Pareto Optimal cuts, and based on structural properties of optimal solutions we devise an improve approach to solve it, thereby improving the original Benders cuts. Computational results demonstrate the efficiency of this approach. 3 - Self Charging Path Problem with Node Visiting Dincer Konur, Missouri University of Science and Technology, 206 Emse, 600 W. 14th Street, Rolla, MO, 65409, United States, konurd@mst.edu In this study, we analyze a shortest path problem with node visiting constraints such that the path should be self-charging. A self-charging path is defined as a path, on which the vehicle traveling charges the energy consumed on the path. This requires using wireless charging and we model integrated design and path decisions. The resulting model is a bi-objective mixed-integer programming model and we propose solution methods to generate Pareto efficient solutions after characterizing important properties of such solutions. A demonstration is also presented. 4 - A Sample Average Approximation Approach for a Network Evacuation Model with Joint Chance Constraints Considering Social Influence Hyeong Suk Na, PhD Candidate, Pennsylvania State University, 234 Leonhard Building, University Park, PA, 16802, United States, hxn144@psu.edu, Necdet Serhat Aybat, Soundar Kumara We consider a network evacuation problem with stochastic traffic demand, where the departure time decision-making behavior of evacuees is affected by social influence. Since it is difficult to characterize the departure behavior due to the inherent uncertainty, we study the evacuation process using the cell transmission model with joint chance constraints (JCCs). To resolve the intractability of JCCs, we adopt a sample average approximation method to handle JCCs. Numerical experiments are implemented to validate the performance of the proposed approximation method. 5 - Packages Delivery Optimization with On-demand Drivers Farzad Daneshgar, University of Maryland, 4324 Rowalt Dr, 302, College Park, MD, 20740, United States, farzadd@umd.edu Nowadays, purchases are made online more than ever and minimizing the delivery time is significantly essential to satisfy customers. Inspired by new innovations such as ride-hailing services, this research aims to develop a mathematical formulation to optimize packages delivery by on-demand drivers. We present a mixed-integer model to determine the scheduling for one-demand drivers by considering elements such as demand volume and operation costs. 6 - Parallel Benders Decomposition Method for Two-stage Stochastic Integer Programs Ragheb Rahmaniani, Universite de Montreal, Pavillon Andre- Aisenst, CIRRELT, P.O. Box 6128, Montreal, QC, H3T.1J4, Canada, Ragheb.rahmaniani@gmail.com, Teodor Gabriel Crainic, Michel Gendreau, Walter Rei In this talk, we discuss parallelization strategies for the Benders decomposition method. The existing parallel variants of the method rely heavily on the synchronization requirement between the master and subproblems. We describe an asynchronous parallel framework for this method. Relaxing the synchronization requirements, however, entails various theoretical and numerical difficulties that will be addressed in this presentation. We present numerical results and provide guidelines regarding how and when to use the proposed parallelization strategies and mention several fruitful research directions.

372B Traffic Management Contributed Session Chair: Yuri Yatsenko, Houston Baptist University, Houston, TX, United States, yyatsenko@hbu.edu 1 - Technology Opportunity Discovery through Extracting Opinion Trigger Based on Word2Vec and Natural Language Processing (NLP) Taeyeoun Roh, First Author, Dongguk University, Seoul, Korea, Republic of, nty92@dongguk.edu, Yujin Jeong, Byungun Yoon Some needs cannot be satisfied in technology same as needs due to there is no specify object that must be satisfied and some object is not related to their technology fields. To solve this problem, this research proposes an approach to discovering new technology opportunity related to opinion trigger(OT). OT is defined as object that users express a feeling. First, OT and their sentimental value are defined through natural language processing and naïve base classifier. Second, OTs and patent keywords which have similar meaning in context are clustered as needs and technology related to needs based on Word2Vec. Finally, technology which can be satisfied needs is derived as technology opportunity. 2 - Investigating Knowledge Management Enablers-a Case Study Hamdy Elwany, Professor of Industrial Engineering and Operations, Alexandria University, 678 El Horreya Ave., Louran, Aexandria, 21532, Egypt, hamdy@elwany.com This study investigates the relationship between Knowledge Management Enablers KMEs and Knowledge Management Success KMS in Middle East and North Africa region MENA industrial companies. Data was collected through an online self-administered questionnaire from 251 respondents in several industries/business sectors in MENA region. The findings of this research proved the existence of the relationship between KMEs and KMS. Also job satisfaction can be considered as KM enabler by having a positive impact on KMS. The organizational culture has a full mediating effect on the job satisfaction and KMS, and a partial mediating effect between organizational structure and IT support. 3 - Technology Adoption in Firm Networks Wei Zhang, University of Hong Kong, Pokfulam Road, Hong Kong, 0050, Hong Kong, zhangw.03@gmail.com We study how firms influence each other when adopting a new technology. Using real data from the semiconductor industry and survival analysis, we show that the chance of a firm adopting a new technology is a U-shaped function of the number of firms that have adopted the technology. In addition, the adoption decision of some firms would positively influence others while some firms would negatively influence others. We use the model to predict the demand of a new technology. 4 - Patent Technical Information Retrieval System Based on Natural Language Processing Wan Wook Ki, Postech, Pohang, Korea, Republic of, wwki86@postech.ac.kr With the development of text mining and natural language processing technology, it became possible to extract desired technical information from the each section of the patents. In this study, we developed a system that can semantically search technical information based on machine learning that can directly help technology planning such as technical function, context, components, purpose, and fusible technology in patents. This enables user to semi-automatically search and analyze patent information, which should have relied on the know-how of existing patent and technology domain experts. 5 - Optimization of Sustainable Capital Lifetime under Environmental Uncertainty Yuri Yatsenko, Professor, Houston Baptist University, 7502 Fondren Road, Houston, TX, 77074, United States, yyatsenko@hbu.edu, Natali Hritonenko Serial replacement of a single capital asset is analyzed under uncertain operating cost and uncertain technological improvements. A modification of classic Economic Life method is suggested and compared to popular asset replacement approach based on real options and dynamic programming. It is shown analytically and numerically that both techniques deliver similar results when the cost volatility is small. The constructed algorithm works equally well for any age- distribution of stochastic operating cost, while the real options technique is developed for geometric and linear cost age-profiles only. The algorithm is illustrated on replacement data for medical imaging devices.

402