Informs Annual Meeting 2017

POSTER SESSION

INFORMS Houston – 2017

50 - Machine Learning for Safer Roads Haridut Athi Shyam Sundar, University of Illinois at Urbana-Champaign, 405 E Stoughton, Apt 20, Champaign, IL, 61820, United States, haridut2@illinois.edu This research primarily focuses on finding interesting and hugely informative patterns in fatal road accidents. Further, it explores how machine learning can be applied to predict fatalities on the roads given the circumstances of a collision. Our work draws inspiration from Los Angeles’ and New York’s ‘Vision Zero’ initiative which aims to achieve zero road fatalities by 2025. Though our initial research is based out of these two city’s public data, our larger objective is to build robust machine learning models that could generalize well to most of the cities. 51 - Sustainable Electricity Markets Mohammad Rasouli, Sustainable Electricity Markets, University of Michigan, 2829 Catalpa Circle, Apt 12, Ann Arbor, MI, 48108, United States, rasouli@umich.edu, Mohammad Rasouli Electricity policy target in majority of the countries is to achieve sustainable and reliable electricity with efficient prices for uncertain future demand.. We provide a model of the industry with the policy targets propose market-based implementations of the electricity policy targets. 52 - Decoding Nonlinear Brain Dynamics from Human EEG Miaolin Fan, Northeastern University, 360 Huntington Avenue, Boston, MA, 02115, United States, fan.mi@husky.neu.edu, Chun-An Chou Decoding EEG activity through pattern recognition helps understand human conscious, perceptions and experiences. However, EEG is a highly complex system driven by nonlinear dynamics. In our study, we characterized the human EEG dynamics based on chaos theory and spectral graph theory. Experiment 1 presents a seizure detection study where graph theoretic method is combined with Recurrence Plots for detecting abnormal state transients using epileptic EEG recordings. In Experiment 2 an emotion recognition task was performed, and we applied RP-based feature extraction method to classify patterns form different emotional states. In conclusion, our method achieved a good accuracy and low latency. 53 - Efficient Algorithms for Some Graph Inverse Problems Jonathan Celaya, Rice University, 6360 S.Main Street, Houston, TX, 77005, United States, jsc7@rice.edu Solving the inverse of a weighted graph problem P entails finding the least modification of weights so that X becomes an optimal solution for P. We present polynomial time algorithms for the inverses of the maximum clique and minimum vertex cover problems on bipartite graphs, and also for the former on block graphs. Both under the 1 and infinity norms. 54 - Understanding Consumers’ Impulse Buying Behavior in Social Commerce Platforms Samira Farivar, PhD Candidate, DeGroote School of Business, DSB A211, DeGroote School of Business, McMaster University, 1280 main street west, Hamilton, ON, l8s 4m4, Canada, farivas@mcmaster.ca, Yufei Yuan Social commerce has identified as a new online commerce platform which enhances users’ information contribution and interaction. Thus, social aspects are the main focus of these environments. Nonetheless, the role of social facets in influencing social commerce users’ behaviors is not yet fully understood. In this study, we are interested in examining social aspects that exist in the social commerce and their influence on users’ buying behavior. Drawing from Latent State-Trait theory, we build our model which suggests that users’ personality traits, environmental cues (social factors), and interactions between these two elements drive users to buy impulsively. 55 - Doubling Supply Chain Profits the Theory of Constraints Way: A Case Study Harshal Lowalekar, Associate Professor, Indian Institute of Management-Indore, Prabandh Shikhar, Rau-Pithampur Road, Indore, 453331, India, lwlherschelle@gmail.com We describe the implementation of Theory of Constraints (TOC) at a large consumer durables company in India. Using TOC’s thinking process the constraints that limited company’s performance in production and distribution were identified and eliminated. TOC’s unique approach helped in drastic reduction of inventories in the supply chain. Shortage and excess of SKUs in distribution almost vanished. Lead times for make-to-order (MTO) items decreased by 75% while the availability of make-to-stock (MTS) items increased to 100%. ROI for the distributors and retailers on the company’s products more than tripled while the profits of the company doubled during the Great Recession of 2008 due to TOC. 56 - Space Efficient Layouts for Block Stacking Warehouses Shahab Derhami, Postdoctoral Fellow, Georgia Institute of Technology, 755 Ferst Drive, NW, Atlanta, GA, 30332, United States, shahab.derhami@isye.gatech.edu In block stacking warehouses, pallets of stock keeping units are stacked on top of one another on the warehouse floor. The arrangement of lanes in the layout of this storage system affects utilization of the storage volume. Existing research that

studies space utilization exclusively attempts to find the optimal lane depth and does not address the design of a space-efficient layout. We describe a model for block stacking warehouses that chooses a set of bay depths and arranges them in a layout to minimize wasted space. We use simulation to evaluate performance of the proposed model through an experimental analysis that covers small to industrial-sized warehouses.

Tuesday Poster Competition Exhibit Hall Tuesday Poster Competition Competition Poster Session

Chair: Sergiy Butenko, Texas A&M University, 4037 Emerging Technologies Building, Mail Stop 3131, College Station, TX, 77843-3131, United States, butenko@tamu.edu Co-Chair: Jeffrey W Herrmann, University of Maryland-College Park, Department of Mechanical Engineering, College Park, MD, 20742, United States, jwh2@umd.edu 1 - A Sequential Game of Defense and Attack on Interdependent Systems Fei He, Texas A&M.University-Kingsville, MSC 191, 700 University Blvd, 700 University Blvd, Kingsville, TX, 78363, United States, fei.he@tamuk.edu, Chiamaka J. Agwuegbo This research studies the defense strategy of an interdependent system in the face of rational attacks. We propose a sequential game between an attacker and a defender on an interdependent system of systems (SoS) to explore the effect of interdependency on the optimal defense strategy. An algorithm of backward induction is developed to determine the equilibrium path of the game. The attacker is the first mover as he applies an attack strategy on constituent systems that maximize his utility. The defender observes and responds by an optimal defense strategy which maximizes her utility. The utility is evaluated as the difference between a player’s reward and the cost of the action. The sensitivity analysis is carried out to compare the effects of different parameters on the attacker and defender’s strategies such as the effectiveness of defense, the effectiveness of attack, the unit cost of defense and the unit cost of attack on constituent systems. 2 - Using Drones to Minimize Delivery Time in Disaster Relief Operations Mohammad Moshref-Javadi, Purdue University, 315 N Grant St., Room 324E, Purdue University, West Lafayette, IN, 47907, United States, moshref@purdue.edu, Lee Seokcheon Drone is an emerging technology which can be used for distribution of products and services. To benefit from advantagesof drones and deal with their limitations, we design a combined system of drones and trucks in which trucks carry bothsupplies and drones and stop at some points in their routes to launch the drones to deliver supplies. The goal is to minimizethe waiting time of the recipients by deciding about the number of drones, launch locations, victims visited by drones, andsequence of visits. The results of the worst-case analysis and case study in Virginia show that this approach can lead tomore than 58% faster delivery. 3 - Ultra-high Precision Predictive Assembly of Composite Fuselage Joins via Surrogate Model Based Control Xiaowei Yue, Georgia Institute of Technology, 755 Ferst Drive NW, ISYE, Atlanta, GA, 30332, United States, xwy@gatech.edu, Yuchen Wen Ultra-high precision predictive assembly is vital for large-scale aircraft production. Passive manual adjustment in current practice is low efficient, non-optimal and experience-dependent. We propose an automated optimal shape control system that can adjust composite parts to an optimal configuration efficiently. The objective is accomplished by (i) building and validating a finite element platform; (ii) developing a surrogate model considering various uncertainties to achieve predictive performance; (iii) conducting feed-forward control to determine the optimal actions. We show the system can significantly reduce assembly time and dramatically improve dimensional quality. 4 - Ranking Alternatives with Partial Information Based on Flexible Interactive Tradeoff Eduarda Frej, Universidade Federal de Pernambuco, Recife, Brazil, eafrej@gmail.com, Adiel de Almeida A new approach for ranking alternatives in multicriteria decision making based on the Flexible and Interactive Tradeoff is presented. FITradeoff is a partial information method for elicitation of criteria weights in additive models. The decision aiding is conducted through a structured elicitation process based on tradeoffs, in which the Decision Maker interactively compares consequences by providing strict preference statements, in such a way that less cognitive effort is spent in the decision making process. An application of a supplier selection problem in a food industry is presented in order to show how this method can be applied to solve real-world cases.

372