Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MA08

4 - On Incentivized-social-inuence-based Programs to Promote Behavioral Changes: A Case Study for Incentivizing Households to Save Energy John Fontecha, University at Buffalo, 413 Bell Hall, Buffalo, NY, 14260, United States, Manjunath Jois, Alexander Nikolaev, Jose Luis Walteros We present a social influence program that allocates investments as random economic incentives for a targeted community to promote changes in its energy- spending behavior. By design, an energy provider funds the program; all the members of the community are eligible to receive an incentive (win); each round of the program features many winners; the winners are chosen randomly but active savers are more likely to win; customers are informed of the winnings in their neighborhoods, fueling the program’s reach. We provide the theoretical basis for the design and operation of such programs and develop methods to optimize their impact by modeling their dependence on the investment allocation strategy. 5 - Workforce Management under Social Link Based Corruption Abhinav Perla, University at Buffalo, 342 Bell Hall, University at Buffalo, North Campus, Buffalo, NY, 14260, United States, Alexander Nikolaev, Eduardo Pasiliao Development of relationships enabling corrupt behavior can be a hindrance to the productivity of an organization. This work introduces the Link Based Corruption (LBC) model, offering a perspective to curb corruptionthrough workforce rotation. The agents are assumed to be embedded into a directed peer-to-peer monitoring network. Corruption is taken to be a threat whenever an agent and their supervisor(s) are all corruption-prone; once they identify each other as such, which takes time, a productivity loss takes place. This work addresses the policy- maker’s problem of fixing the agent monitoring structure and timing the workforce rotation so as to minimize the expected long-time loss. n MA08 North Bldg 124A Infrastructure Optimization Sponsored: Optimization/Network Optimization Sponsored Session Chair: Robert Mark Curry, Clemson University, Clemson, SC, 29634, United States 1 - Robust Location of Transparent Interdictions on a Shortest Path Network N. Orkun Baycik, Shenandoah University, Winchester, VA, United States, Kelly Sullivan We study a shortest path network interdiction problem in which the follower seeks to find a path of minimum length on a network and the leader seeks to maximize the follower’s path length by interdicting arcs. We consider placement of interdictions that are not visible to the follower; however, we seek to locate interdictions in a manner that is robust against the possibility that some information about the interdictions becomes known to the follower. We formulate the problem as a bi-level program, and apply a Benders decomposition approach to optimally solve it. We derive supervalid inequalities to improve the performance of the algorithm and test the algorithm on grid networks and acyclic graphs. 2 - Two-stage Stochastic Interdependent Networks Mitigation and Restoration Cheng-Lung Chen, University of Central Florida, Orlando, FL, United States, Qipeng Zheng, Alexander Veremyev, Vladimir Boginski We propose a two-stage stochastic program for determining the optimal mitigation and restoration strategy on two-layer interdependent networks with cascading node failure caused by random disruption. Previous studies mainly focused on either static or deterministic type of network failure, while our model features a dynamic failure propagation given initial stochastic node disruption. The first-stage problem allows capacity expansion and node fortification to mitigate the impact of disruption, while the objective of second-stage is to restore network performance with minimum costs. We propose a delayed cut generation algorithm with various types of cutting plane to solve the problem. 3 - A Post Disaster Infrastructure Restoration Model: The Interdependent Integrated Network Design and Scheduling Problems with Machine Movement. Aniela Garay Sianca, University of Arkansas, Fayetteville, AR, 72701, United States, Sarah G. Nurre Over the past years, post-disaster operations research models have become necessary due to the increasing number of disasters. We propose a new optimization model that looks to schedule work crews to restore damaged components on a set of interdependent networks while explicitly considering the availability and restoration of the transportation network which enables work crew movement. On realistic data representing interdependent infrastructure networks in Panama, we perform computational experiments to deduce insights and show improvement as compared to existing restoration models.

n MA06 North Bldg 122C Decomposition in Large-Scale Computational Optimization Sponsored: Optimization/Computational Optimization and Software Sponsored Session Chair: Matthew Galati, SAS Institute, Inc., Glen Mills, PA, 19342, United States 1 - Decomposition Algorithm for a Capacity Expansion Problem with Permanent and Temporary Capacities Rahman Khorramfar, PhD Student, NC state University, 2804 Brigadoon Dr, 1, Raleigh, NC, 27606, United States, Osman Ozaltin We study a type of stochastic capacity expansion problem where there are multiple capacities to open in each planning period. We consider capacities of temporary nature and permanent capacities, and formulated the model as a multi-stage mixed integer stochastic program with uncertain demand. We apply Dantzig-Wolfe decomposition algorithm to solve the problem. We also extended our model to include a mean-risk objective in order to reflect the sensitivity of decision makers to risk factors. Computational results are presented for different applications. 2 - Dantzig Wolfe Decomposition Approach to Inland Waterway Disruption Liliana Delgado Hidalgo, Universidad del Valle, Cl. 13 #100-00, Cali, 760001, Colombia We study inland waterway disruption response to redirect disrupted barges to available terminals and prioritize offloading to minimize total cargo value loss. We formulate a mixed integer linear programming model as a heterogeneous vehicle routing problem with time windows (HVRPTW) and working to reformulate the HVRPTW via Dantzig-Wolfe decomposition approach to explore advantages of the Branch-and-Price and column generation approaches. A case example is presented to illustrate our approach and findings. n MA07 North Bldg 123 Optimizing over Networks with Dynamics Sponsored: Optimization/Network Optimization Sponsored Session Chair: Christopher Quinn, Purdue University, West Lafayette, IN, 47907, United States 1 - Optimal Versus Selfish Resource Allocation Dynamics Over Networks S. Rasoul Etesami, University of Illinois at Urbana-Champaign, Urbana, IL, United States Motivated by many emerging resource allocation problems such web caches and peer-to-peer networks, we consider and study optimal and selfish resource allocation over a network of agents. In this model the nodes can store only a limited number of resources while access the remaining ones through their neighbors. We first study the complexity and approximability of the optimal resource allocation. We then extend our results to selfish setting by formulating the problem as a noncooperative game. We show that such a game always admits a pure-strategy Nash equilibrium and device stochastic learning dynamics to obtain one of such equilibrium points. 2 - Dynamic Public Learning in Networks of Strategic Agents AMIR Ajorlou, Massachusetts Institute of Technology, 32 Vassar Street, 32-D569, Cambridge, MA, 02139, United States, Ali Jadbabaie There are abundant sources of information (e.g., news channels, social networks, online forums, surveys, and data bases), with every person having access to a personalized set of the sources. The aggregate action of the whole population partially accumulates the information shared within different information communities. An interesting question here is how the quality of information aggregation is affected by the influences between and within different information communities. We show that the inter-community interactions exhibit a double-edged effect that may result in a phase transition in quality of learning between different information-action structures. 3 - Sparse Modeling of Network Interactions Christopher Quinn, Purdue University, 315 N. Grant Street, West Lafayette, IN, 47907, United States There are many natural and emergent systems with complex interactions between components. It is often easier to conduct observational than experimental studies. However, there are significant statistical and computational challenges in modeling and characterizing the interactions from observational data. In this talk, we will discuss recent methods to efficiently and non- parametrically identify sparse network approximations from data, with guarantees of optimality under certain conditions.

121