Informs Annual Meeting 2017

WB09

INFORMS Houston – 2017

4 - Predicting Stock Market Short-term Price Based on Machine Learning

positioning and pricing of the brands. We find that, when the store brand introduction is not very costly, the retailer voluntarily shares information with the manufacturer so that the both product are better positioned even though this increases the competition between the brands. 4 - Supply Chain Expansion and Integration C. Gizem Korpeoglu, University College London, Gower Street, London, WC1E 6BT, United Kingdom, c.korpeoglu@ucl.ac.uk, Ersin Korpeoglu, Soo-Haeng Cho We study expansion and integration of supply chains where multiple suppliers sell to multiple retailers. Though retailers influence wholesale price in practice, prior literature fails to capture this influence. To overcome this limitation, we develop a novel model that considers retailers’ buyer power. We show that more retailers raise wholesale price and can improve retailer profits, and that supply chain expansion is more beneficial than prior literature finds. We then show that integration of two supply chains may reduce total profits in a retailer-oriented supply chain. This result suggests that as UK disintegrates from EU via Brexit, supply chain profits may increase in UK and decrease in EU. 330B Auctions/Mechanism Design Contributed Session Chair: Wenwei Wang, Hong Kong University of Science & Technology, Kowloon, Hong Kong, wwangaw@connect.ust.hk 1 - Creating a Transition Schedule for Broadcast Television Karla L.Hoffman, George Mason University, System Eng. and Operations Research Dept, 4400 University Drive Mailstop 4a6, Fairfax, VA, 22030, United States, khoffman@gmu.edu Due to the successful Broadcast Incentive Auction, many television stations in the United States and Canada will be required to move to new channels. In many cases, one station will not be able to move to its new channel until another station vacates its old channel. This difficulty led the FCC to create a phased transition plan to ensure that television stations maintain coverage during the transition while completing the entire transition in 39 months. In this talk, we will present the problems presented by the transition and the methodology that the FCC used to resolve these issues. 2 - Sequential Effects and Bidder Heterogeneity in B2B Auctions We study a longitudinal dataset of truckload spot auctions in which we observe exogenously-imposed business constraints on the bidding firms - some firms can perform work only using assets that they own and operate, while others must outsource work - i.e., brokers. These business constraints manifest themselves in strikingly different behavior by the bidding firms in single and sequential auctions. Brokers adjust their prices in subsequent auctions based on whether they won or lost the previous auction, while firms who operate assets decide whether or not to place a bid. We examine other sources of heterogeneity and discuss implications for our focal industry. 3 - Deterministic Truthful Mechanisms for Interval Scheduling with Applications to Cloud Computing Peter McGlaughlin, Graduate Student, University of Illinois at Urbana-Champaign, 117 Transportation Building, 104 S. Mathews Ave, Urbana, IL, 61801, United States, mcglghl2@illinois.edu, Jugal Garg Motivated by cloud computing, we study a market-based approach for scheduling where jobs have hard deadlines and prefer earlier completion times. In our model, completing a job earns its present value. Users submit job requirements to the cloud provider who non-preemptively schedules jobs to maximize social welfare, i.e. the present value of completed jobs. Using a greedy algorithm, we obtain a 1/(1 -b) approximation to the optimal schedule, where b < 1 is the discount factor shared by all jobs. For sufficiently small b, this beats the best known approximation algorithms applicable to the problem. Finally, we construct a pricing rule to incentivize users to truthfully report all job requirements. 4 - Mechanism Design with Efficiency and Equality Wenwei Wang, Hong Kong University of Science & Technology, Room 5569, Academic Building, HKUST, Kowloon, Hong Kong, wwangaw@connect.ust.hk, Zhou Chen, Qi Qi, Changjun Wang In this work, we consider the problem of allocating a set of homogeneous goods among multiple strategic players to balance the efficiency and equality from a game-theoretic perspective. We develop a general truthful mechanism framework which optimally maximizes the resource holder’s efficiency while guaranteeing some equality constraints. We use the Generalized Gini index as the equality measure, and fully characterize the optimal allocations. Based on the characteristics, we propose a polynomial-time algorithm that computes the optimal allocation and corresponding truthful payments for any bids profile. We also show how the optimal efficiency varies under varying equality constraints. WB10 Alex Scott, Assistant Professor, Northeastern University, 216 Hayden Hall, Boston, MA, 02038, United States, f.scott@northeastern.edu

Xing Wang, Auburn Univeristy, 371 W Glenn Avenue, Apt 14, Auburn, AL, 36830, United States, xzw0005@auburn.edu, Bin Weng, Lin Lu, Fadel Mounir Megahed, Alexander Vinel New forms of collective intelligent have emerged with the rise of the Internet. The changes on these platforms will significantly affect the stock market. In this study, we combined disparate online data sources with the traditional time series signal. Several machine learning techniques are applied to train a short-term price prediction model, including decision tree, support vector machine, boosting, as well as different neural network architectures in deep learning. The internal nature and applicability of these models would be analyzed, and their performances would be compared. 5 - Conducting Confirmatory Factory Analysis on an Instrument to Measure Systems Thinking Capacity Morteza Nagahi, PhD Student, Graduate Research Assistant, Mississippi State University, 450 Campus Trails, Apt 229B, Starkville, MS, 39759, United States, mn852@msstate.edu The Instrument to Assess Capacity for Systems Thinking was developed to measure individual systems thinking capacity, providing a useful baseline to facilitate agent-assignments, training, and education for effective complex systems problems. The research-based instrument generates systems thinking profiles by capturing the state of individual inclination toward systemic thinking as well as a preference for engaging and solving complex system problems. Researchers perform confirmatory factor analysis (CFA)to verify the factor structure of the construct and evaluate construct validity. 6 - Bilevel Programs for Cross Validation in Classification Models Statistical methods that utilize cross validation for model selection are shown to be effective in the classification process while identifying the most important features that influence the classification. In this work, we propose a bilevel framework to be used in the cross validation procedure and develop a novel genetic algorithm as a solution approach. We implement the proposed framework and solution procedure in an influenza classification case study where we classify the antigenic variation between influenza virus strains. Our results demonstrate that bilevel programming can improve the strength of classification models when implemented in the cross validation process. 330A Supply Chain Economics Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: C Gizem Korpeoglu, University College London, London, WC1E 6BT, United Kingdom, c.korpeoglu@ucl.ac.uk 1 - Supplier Diversification under Random Yield: the Impact of Price Postponement Nan Yang, University of Miami, Coral Gables, FL, United States, nyang@bus.miami.edu, Lingxiu Dong, Guang Xiao We characterize a firm’s pricing and sourcing decisions under supply yield uncertainty, and study the interplay between supply diversification and pricing. We compare the optimal sourcing decisions under two pricing schemes: ex ante pricing and responsive pricing. Although the firm is better off with responsive pricing, the effect of price postponement on the optimal sourcing decision and the firm’s need for supply diversification varies and is influenced by factors such as unit procurement cost, supply portfolio, and supply reliability. 2 - Inventory Strategy of Omni-channel Xiaolin Xu, Nanjing University, School of Business, 23 Hankou Road, Nanjing, 210093, China, xuxl@nju.edu.cn, Ming Hu, Weili Xue, Yi Yang The future of retailing seems to be omnichannel as the mix of online vs. offline selling. Online retailers such as Amazon move offline by setting up local stores, whereas brick-and-mortar retailers such as Walmart and Target have long set up online stores. We show that buy-online, pick-up-in-store strategy has a drastically different impact on retailers of different natures. 3 - Consumer Taste Uncertainty in the Context of Store Brand and National Brand Competition Arcan Nalca, Smith School of Business, Queen’s University, Goodes Hall Queen’s University, Kingston, ON, K7L.3N6, Canada, arcan.nalca@queensu.ca, Tamer Boyaci, Saibal Ray We study how a retailer can benefit from acquiring information about uncertain consumer tastes in the presence of competition between the retailer’s store brand and a manufacturer’s national brand. We identify retailer’s optimal information sharing strategy with the manufacturer as well as the equilibrium product Joseph Kapena Agor, PhD Student, North Carolina State University, 2500 Stinson Drive, Apt C, Raleigh, NC, 27695, United States, jkagor@ncsu.edu WB09

476