Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TD52

6 - Improving Patient Flow Metrics at Outpatient Clinics through Effective Scheduling by Modeling Two Stage Mean Risk Stochastic Programming Samira Fazel Anvaryazdi, Louisiana Tech University, Ruston, LA, United States, Saravanan Venkatachalam, Ratna Babu Chinnam We discuss methods for improving patient flow metrics at outpatient clinics through effective appointment scheduling policies by applying risk-averse two- stage stochastic mixed-integer linear programming. We improve patient flow metrics: direct and indirect wait time, provider’s overtime and idle time considering patient’s no-show behavior, stochastic server, follow-up surgery appointments, and overbooking. The aim is to increase throughput per session while providing timely care, continuity of care, and overall patient satisfaction as well as equity of resource utilization. n TD52 North Bldg 231C Flash Session II General Session Chair: INFORMS, 5521 Research Park Drive, Suite 200, Catonsville, MD, 21228, United States 1 - Developing a Model to Select the Best Multiple Sites to Install Wind Farms Nazanin Naderi, Youngstown State University, Youngstown, OH, 32606, United States, Shabnam Rezapour Renewable energy production has gained popularity worldwide due to the rapid depletion of fossil fuels and a global effort to reduce carbon emissions to mitigate climate change. Because of the formidable costs associated with wind energy development, the locations for new wind turbines need to be carefully selected to provide the greatest benefit for a given investment. This paper primarily focus on developing a model to find the best sites among alternative sites, could help find the efficient wind resources and lands that are well suited for wind turbines. The model developed in this paper compares combined sites based on nine diff erent criteria and prioritizes them. 2 - Robust Service Network Design for Disaster Relief Syed Tariq, Lahore University of Management Sciences (LUMS), Shah House, A-10, Block 2, Chapal Sun City, Lahore, 75280, Pakistan,, Muhammad Naiman Jalil The service network design problem we discuss is concerned with facility location and prepositioning of relief goods in the preparedness stage and the reliability of network performance during response. We argue that the location of facilities determines the possible damage to facility capacities, access to vulnerable populations in their time of need, and timeliness of resupplying the facility in the response phase. In addition to facility and network related uncertainties, we evaluate the service network design in terms of both deprivation and logistics cost when some locations are more prone to disaster-related damage than others. 3 - Optimizing the Redundancy of IoT Networks Mohammad Askar Saoud, Kuwait University, P.O. Box 54, Shamiya, 71661, Kuwait, Hamid Al-Hamadi In this paper, we aim to maximize the lifetime of IoT networks using a redundancy management approach. As such networks are set up, paths must be determined for the IoT devices. While increasing the number of paths (which reflects redundancy level) increases reliability, energy consumption also increases. Thus, one needs to choose paths such that proper reliability level is maintained and simultaneously energy consumption is minimized. So, we propose an INLP model to represent the problem, and our objective is to maximize the Mean Time to Failure (MTTF). 4 - Parking Reservation Disturbances Su Xiu Xu, Department Head, Jinan University, 206 Qianshan Road, Zhuhai, 519070, China This paper considers an auction-based parking reservation problem where a platform is the auctioneer and drivers are bidders. The platform manages multiple homogenous parking spaces. A winner may leave earlier or occupy the parking space longer than the time he has reserved. The phenomena are called (ex post) demand disturbances, which can only occur after the last auction terminates. The platform may punish or compensate a driver who causes demand disturbance. We investigate three types of driver behaviors: gain/loss neutrality, loss aversion, and gain seeking, and examine the reference effects. We propose an effective multi- stage Vickrey-Clarke-Groves auction mechanism.

5 - Inventory-routing Problem with Pick-up and Deliveries: Formulation, Solving Approach, and Sustainability Perspectives Claudio Sterle, University “Federico II” of Naples, Via Claudio 21,

Naples, 80125, Italy, Grazia Speranza, Claudia Archetti, Maurizio Boccia, Sforza Antonio, Maria Elena Nenni

The Inventory Routing Problem (IRP) consists in the distribution of one or more products from a supplier to a set of customers over a discrete planning horizon. In this work we address the IRP with Pickups and deliveries where a single commodity has to be picked up from pickup customers and delivered to delivery customers. A fleet of vehicles is available, starting and ending their routes at the supplier’s depot. The objective is to determine a distribution and collection plan minimizing routing and inventory costs. We propose a flow formulation and a branch-and-cut algorithm. Computational results on benchmark instances show that the algorithm outperforms state-of-the-art methods. 6 - Agent Scheduling for Call Center Operation Ayush Patnaik, IIT Kharagpur, A-318, VS Hall, IIT Kharagpur, Kharagpur, 721302, India, Swapnil Waghmare, Samik Raychaudhuri Agent scheduling, including scheduling breaks, is an important problem for call center operations. In this presentation, we present an optimization model that aims to provide us with the best possible schedule while satisfying numerous constraints such as operational hours, breaks, no transportation etc. It gives us the optimal allocation of agents in each shift so as to minimize agent under-allocation for the forecasted chats. We use the result of this model in a follow-up model, which allocates breaks of different types e.g., training breaks, lunch breaks etc, to individual agents. 7 - A Drift-variance Diffusion Control Model for New Venture Creation Zhengli Wang, Stanford University, 655 Knight Management Way, 655, Stanford, CA, 94305, United States, Stefanos Zenios We model new venture creation as a stochastic control process in which the value of the venture evolves as a diffusion process, and the drift and variance is controlled by the entrepreneur. The venture goes bankrupt when the diffusion process hits a lower boundary and is successful with a substantial monetary reward when the process hits an upper boundary. The entrepreneur incurs cost for the controls she selects and her objective is to maximize her total cumulative infinite horizon reward. We analyze the situation where the entrepreneur has two control options and we demonstrate that the optimal policy is at most a two- interval policy. 8 - Addressing Orientation-symmetry in the Time Window Assignment Vehicle Routing Problem Kevin Dalmeijer, Erasmus University Rotterdam, P.O. Box 1738, Tinbergen Building H8, Rotterdam, 3000 DR, Netherlands, Guy Desaulniers The Time Window Assignment Vehicle Routing Problem (TWAVRP) is the problem of assigning time windows for delivery before demand volume becomes known. For TWAVRP instances that are difficult to solve by current methods, we observe many similar solutions in which one or more routes have a different orientation, i.e., the clients are visited in the reverse order. We introduce a new branching method that eliminates this orientation-symmetry from the search tree, and we present enhancements to make this method efficient in practice. Through computational experiments, we show that our algorithm outperforms other solution methods, and we solve 29 previously unsolved benchmark instances to optimality. 9 - Patrolling Security Games with Mobile Adversaries and Entrance Nodes Zhifan Xu, Rutgers University, 18M Reading Road, Edison, NJ, 08817, United States Patrolling is an important operational decision when safeguarding a public area against adversarial attack. In this paper, we proposed an infinite horizon Stackelberg security game on a graph where the defender commits to a patrolling strategy first. The attacker decides his intruding route by solving a Markov decision process given patroller’s movement probabilities and location in each turn. We also introduced the concept of detection probability for each nodes in the graph to model the difficulties of searching different area. To find the Stackelberg equilibrium, we proposed a bi-level programming model that can be solved using bilinear programming and global optimization techniques. 10 - Deep Learning Applications in Financial Services Trevor Kennedy, Ally Financial, Philadelphia, PA, United States Deep learning, newest incarnation of machine learning is just beginning to be applied in industry. We present a case study of several applications of deep neural networks in the financial services industry including best practices and limitations of this approach.

365