Informs Annual Meeting 2017

TC45

INFORMS Houston – 2017

TC45

procurement, we need to bound certain online scheduling problems. Here, we prove that the DSPT algorithm is 2-competitive for the online version of minimizing modified total tardiness. 2 - Preemptive Parallel-machine Scheduling Problem of Maximizing the Number of On-time Jobs Marc E. Posner, Ohio State University, 296 Baker Systems Engineering, 1971 Neil Avenue, Columbus, OH, 43210-1271, United States, posner.1@OSU.edu, Hui-Chih Hung, Bertrand M.T. Lin, Jun-Min Wei We study the classical preemptive parallel-machine scheduling problem of maximizing number of on-time jobs. While known to be NP-hard, no theoretical analysis of approximation algorithms exists. We propose heuristics based on different design strategies, which have asymptotically tight relative errors of 1/2. Experimental tests evaluate computational performance. As part of the analysis, a new non-standard mixed integer formulation is developed. 3 - Cutting-plane Methods for Resource-constrained Project Scheduling Problems in Strategic Mine-planning Alessandro Hill, Universidad Adolfo Ibanez, Santiago, Chile, alessandro.hill@uai.cl, Marcos Goycoolea We consider resource-constrained project scheduling problems which are powerful models for mine planning. These notoriously hard problems aim at scheduling given tasks under bounded availability of multiple resources per time period, while respecting given precedence relations, in order to minimize the total project duration or maximizing a profit function. We present novel polyhedral techniques based on various problem-specific cuts. In our mathematical programming framework, we borrow fast enumeration subroutines from constraint programming. We show the detailed potential of our approaches within branch and cut algorithms for literature instances as well as real-world problems. 4 - Activity Scheduling in Mine Construction Valentina Rojas, Universidad de Chile, Santiago, Chile, vjrojasc@gmail.com, Rodrigo A. Carrasco, Alessandro Hill, Nelson Morales Construction of a mine is critical because it defines the available area for production, but proper model of a plan schedule can be hard, because there may be different ways to reach certain part of the mine, different potential technologies to use or because some activities may collaborate in order to achieve a common goal. In this talk we present a MIP model that abstracts these considerations and compare solution approaches on several mining methods. 360F Leveraging Online Platforms to Manage Services Sponsored: EBusiness Sponsored Session Chair: Sreekumar R. Bhaskaran, Southern Methodist University, Dallas, TX, 75275, United States, sbhaskar@mail.cox.smu.edu 1 - Leveraging Online Platforms to Manage Services Sanjiv Erat, University of California-San Diego, Rady Sch. Mgmt. Otterson Hall, 9500 Gilman Drive MC 0553, La Jolla, CA, 92093-0553, United States, serat@ucsd.edu Many large firms are increasingly relying on open innovation, often requesting their own customers to brainstorm and come up with new products and services. While firms are becoming more adept at using IT to collect ideas from outside the firm boundaries, the evaluation and filtering of ideas is still a costly and labor intensive process. This talk shall discuss how some recent innovations in natural language processing can be effectively leveraged to automate the evaluation and filtering of large idea pools, thus making open innovation accessible and affordable even for smaller firms. 2 - Strategies of Coupon Issuance in Service Industries Simin Li, Kellogg School of Management, Northwestern University, 2211 Campus Drive, Evanston, IL, 60208, United States, simin.li@kellogg.northwestern.edu, Kejia Hu, Martin Lariviere We empirically study how service providers strategically design coupons, based on marginal service cost (MSC), to facilitate demand-supply coordination during rush hours. We find service providers with higher MSC launch coupons earlier with less discount. We confirm strategic coupon issuance effectively shift demand of experienced coupon users. Via counterfactual analysis, we propose a strategy tailoring to MSC and customer responsiveness. TC47

360D Revenue Management Contributed Session Chair: Peter McGlaughlin, University of Illinois at Urbana-Champaign, Urbana, IL, United States, mcglghl2@illinois.edu 1 - The Interaction between Product Rollover Strategy and Pricing Scheme JingChen Liu, Peking University, Yiheyuan Road 5th, Haidian District, Beijing, 100871, China, liujingchen@pku.edu.cn, Xin Zhai, Lihua Chen This study proposes a two-period model incorporating forward-looking consumers and market heterogeneity to explore the interaction between rollover strategy and pricing scheme of an innovating firm offering trade-in program. The firm uses either single or dual rollover to launch new generation, and chooses from price skimming and penetration pricing for consumer segment. We find that the optimal pricing scheme and rollover strategy depends on the innovation increment of new product, the salvage value of old product and how strategic consumers are. Moreover, the firm can be worse off deploying trade-in especially when innovation increment is not sufficient or consumers are relatively myopic. 2 - Product Recall through Group-buying with out Deadline JingChen Liu, Peking University, Yiheyuan Road 5th, Haidian District, Beijing, 100871, China, liujingchen@pku.edu.cn, Ming Hu, Xin Zhai This study considers a monopolistic firm selling one contemporary product to consumers with heterogeneous valuation who arrive according to a Poisson process. In order to better price discriminate between high- and low-end consumers, the firm can set up a product recall program, under which the firm will offer the recalled product to consumers who join if the sign-up number reaches the pre-announced threshold. We find that product recall program can perform as a win-win solution, namely the firm can improve its own profit without harming the consumers. 3 - An Analysis of Buy X Get 1 Free Reward Programs Yan Liu, Tianjin University, Room C1211, The 25th Building, Tianjin University, No. 92 Weijin Road, Tianjin, 300072, China, hnnyly@gmail.com, Dan Zhang ”Buy x, get 1 free” reward programs are widely observed in practice. We model such programs as an infinite horizon stochastic game between a monopolistic seller and a population of customers with different product valuation and shopping frequencies. We characterize conditions under which the programs are profitable and discuss their welfare implications. Moreover, we show that “buy 1, get 1 free” program provides a tight bound on seller revenue and is optimal under certain conditions. 4 - Reinforcement Learning for Network Revenue Management Peter McGlaughlin, Graduate Student, University of Illinois at Urbana-Champaign, 409 W. Elm St, Apt 12, Urbana, IL, 61801, United States, mcglghl2@illinois.edu Network Revenue Management concerns the optimal allocation of a fixed set of resources across various segments of market demand. Applications require heuristic solutions as the underlying optimization problem is intractable. We propose reinforcement learning as an alternative. Recently, these algorithms enjoyed widely popularized success in solving large, challenging sequential decision making problems. Thus, reinforcement learning seems a natural candidate for a new approach to network revenue management. Numerical experiments show significant revenue increases, about 10%, over existing heuristics. 360E Scheduling with Applications Invited: Project Management and Scheduling Invited Session Chair: Rodrigo A. Carrasco, Ph.D., Universidad Adolfo Ibáñez, Santiago, 7941169, Chile, rax@uai.cl 1 - Incorporating Suppliers into Capital Project Scheduling Arman Jabbari, University of California-Berkeley, 2012 Del Norte, Berkeley, CA, 94707, United States, a.jabbari@berkeley.edu, Philip Kaminsky The majority of capital projects are completed after their intended completion date and over budget, often due to inefficiencies in the way their supply chains are managed. After conducting extensive with practitioners in the oil and gas industry, we have identified the relationship between the project schedule and the procurement plan as one cause of these problems. As part of a larger project to develop approaches to simultaneously develop schedules and plan TC46

352