Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TE70

2 - The Watermelon Algorithm for The Bilevel Integer Linear Programming Problem Lizhi Wang, Iowa State University, IMSE and ECPE, Ames, IA, 50011, United States We will review the watermelon algorithm for the bilevel integer linear programming problem and present an improved version of the algorithm that uses a number of techniques specifically designed for the bilevel structure of the problem to improve the efficiency of the algorithm. 3 - Lagrangian Dual Decision Rules for Multistage Stochastic Integer Programming Merve Bodur, University of Toronto, 5 King’s College Road, Toronto, ON, M5S 3G8, Canada, James Luedtke, Maryam Daryalal We propose Lagrangian dual decision rules that yield a new approximation approach for multi-stage stochastic integer programming problems. We investigate techniques for using these decision rules to obtain bounds on the optimal value and a primal feasible policy; and compare the strength of the relaxation from different techniques. Numerical results will be presented to illustrate the quality of the obtained bounds and policies. 4 - Representability and Solvability of Mixed-integer Bilevel Programming Problems Sriram Sankaranarayanan, Johns Hopkins University, Baltimore, MD, 21218, United States, Amitabh Basu, Christopher Ryan We show that the feasible regions modeled by continuous bilevel constraints, complementarity constraints and polyhedral reverse convex constraints are the same. We show that their feasible regions model the finite union of polyhedra. We contrast it with the modeling power of mixed-integer problems that model finite union of polytopes along with an integer cone. Then we show that allowing both integer constraints and continuous bilevel constraints, one can model finite union of polyhedra and finite union of integer cones. Having said that, by fixing certain variables, we show non-emptyness of a relaxed mixed-integer bilevel set can be decided in polynomial time. Chair: Debdatta Sinha Roy, Robert H. Smith School of Business, University of Maryland, College Park, MD, 20742, United States 1 - An Improved Quasi-experiment to Evaluate the Effectiveness of Performance-based Reward Lingfeng Li, Vanguard Group, Wayne, PA, United States Performance-based reward (PBR) has been adopted to facilitate the achievement of organizational goals. However, evidence on the effectiveness of PBR is limited, and methods to evaluate PBR effectiveness have not been well established. This work is the first effort to introduce an improved quasi-experiment (IQE) to obtain high level of evidence to support decision-making on the application of a PBR. Our pilot results suggest the effectiveness of a PBR could vary across operational settings: PBR could improve the productivity of employees who process not-in- good-order items by at least 5% (p-value = 0.00005). However, evidence is insufficient for employees who process in-good-order items. 2 - Portfolio Choice of Renewable Energy Chen Wei, University of Electronic Science and Technology of China, Chengdu, China, Ma YongKai, Tang Xiao Wo Most countries used the Energy Internet,which has a key feature that the electricity is mainly generated by renewable energy, to reduce carbon emissions. Renewable energy (i.e., wind and solar) is volatility, however, renewable energy combination on-grid can reduce the renewable energy volatility. This paper focuses on the portfolio choice of renewable energy. We constructed a model that utility firms, including conventional and renewable energy firms, supply the electricity to the market, in different pricing policies. We find that flat pricing leads to a higher investment level for solar energy, meanwhile, peak pricing leads to invest more wind energy. 3 - Sequencing Mixed Models on an Assembly Line with Variable Rate Launching and Open Stations Tobias Moench, WHU-Otto Beisheim School of Management, Vallendar, Germany, Arnd H. Huchzermeier Introduction of new models on an assembly line pools demand risk, but also increases costs of idle time, utility work and space. Opening up station boundaries and introducing variable rate launching instead of a fixed tact time reduces all costs simultaneously. Moreover, the overall line length is reduced significantly. n TE72 West Bldg 211A Practice- Operations Management I Contributed Session

n TE70 West Bldg 106B

Joint Session DEA/Practice Curated: Applications III Emerging Topic: Productivity, Efficiency and Data Envelopment Analysis Emerging Topic Session Chair: Xiang Ji, University of Science and Technology of China, 800 W Renner Road, Richardson, TX, 75080, United States 1 - A Heterogeneous DEA-based Eco-efficiency Measurement of China’s Industrial Sectors Mingjun Li, University of Science and Technology of China, Hefei, China, Jie Wu Environmental problems brought by industry are attracting extensive attention so a comprehensive analysis of industrial environmental performance is increasingly important. In this paper, we extend the DEA model to consider two-sided non- homogeneous problems, handling DMU sets that have non-homogeneity in both inputs and outputs. This is different from the previous researches which generally focus on regional data to avoid heterogeneity. With this more realistic analysis of environmental efficiency, the Chinese government can make more informed decisions to realize sustainable industrial development. 2 - DEA Cross-efficiency Model with Reciprocal Behavior Feng Li, Southwestern University of Finance and Economics, Chengdu, 611130, China A reciprocal behavior is responding much nicer to friendly actions and much nastier to hostile actions. This paper integrates the reciprocal behavior among Decision Making Units (DMUs) into Data Envelopment Analysis (DEA) cross- efficiency approaches. To this end, a series of games are developed by viewing each DMU as a player and adopting different evaluation strategies. Several equilibrium points are obtained for different strategies, and game theoretical implications are provided. Finally, the proposed approach is applied to logistics industries. 3 - Using A Hybrid Heterogeneous DEA Method to Benchmark China’s Sustainable Urbanization Xiang Ji, University of Science and Technology of China, 96 Jinzhai Rd, Hefei, 230026, China, Jie Wu, Qingyuan Zhu, Jiasen Sun Due to recent official policy changes, China’s sustainable urbanization has entered a new type of multi-target heterogeneous situation which has never been scientifically researched before. To deal with this new heterogeneous scenario, we introduce a hybrid heterogeneous data envelopment analysis (DEA) method that contains input segment estimation and efficient frontier construction. We also introduce a bi-level benchmarking method that benchmarks each city’s eco- efficiency and uses those benchmarks to guide sustainable urbanization completion in an empirical study of 206 Chinese prefecture-level cities’ sustainable urbanization. n TE71 West Bldg 106C Joint Session ICS/IOS Discreet: Advances and Applications in Multilevel Mixed-integer Optimization Sponsored: Computing Sponsored Session Chair: Leonardo Lozano, University of Cincinnati, Cincinnati, OH, 45219, United States Co-Chair: Christopher Ryan, University of Chicago, Chicago, IL, 60637, United States 1 - Robust Optimization with Mixed-integer Uncertainty Sets Juan Sebastian Borrero, Assistant Professor, Oklahoma State University, 322 Engineering North, Stillwater, OK, 74078, United States, Leonardo Lozano We consider robust linear optimization problems where the uncertainty set is a mixed-integer linear feasibility region. Such a class of sets allows us to model complex uncertain phenomena where there are logical, combinatorial, or disjunctive relationships in the uncertain data. We show that optimizing over this type of sets greatly reduce the conservativeness of the optimal solutions when compared to a convex counterpart. However, the reduction comes at the price of increasing the computational complexity of the problem. In particular, the problem is NP-hard and we develop iterative solution methods based on decomposition techniques that solve many instances of the problem in a reasonable time.

404