Informs Annual Meeting 2017

WA46

INFORMS Houston – 2017

WA45

the visual similarity. This work is done with an academic-industry collaboration and the method has been tested and validated with actual data.

360D Retail Management Contributed Session Chair: Shin Woong Sung, KAIST, Daejeon, United States, sw.sung@kaist.ac.kr

WA46

360E Project Management Contributed Session Chair: Sinan Obaidat, University of Arkansas, Fayetteville, AR, United States, sfobaida@uark.edu 1 - Efficient Reproduction Scheduling Against Defectives under Uncertainty There are many uncertain factors such as defectives and machine failure in production process. We have to deal with them by rescheduling or adaptive scheduling, which may cause the delivery delay and excess inventory. Robust schedule for uncertainty brings us the reduction of delivery delay and inventory. We consider to provide the appropriate production plan by taking into account of the uncertainty. 2 - Production Planning and Quality Testing under Random Yield Candace Arai Yano, University of California-Berkeley, IEOR.Department, 4141 Etcheverry Hall, Berkeley, CA, 94720- 1777, United States, yano@ieor.berkeley.edu, Sandra Transchel We study a co-production system in which there is a two-stage testing process with limited capacity. Finished products are differentiated according to their quality levels in two dimensions, and the firm sells these products at different prices. Demand can be fulfilled using a substitute product of equal or higher quality but at the price of the requested quality. We develop a stochastic dynamic programming model with three types of decisions: (i) ordering decisions, (ii) testing policy, and (iii) allocation of the post-final-test inventory to the quality- differentiated demands. 3 - Performance Analysis of Multi-product Manufacturing Systems with Dedicated Finite Buffers Kyungsu Park, University of Wisconsin-Madison, Madison, WI, United States, kyungsu@kaist.ac.kr, Jingshan Li Many flexible manufacturing systems process multiple products with dedicated buffers. In such systems, scheduling and control polices play a critical role. Priority, cyclic, and WIP-based policies are among the most typical ones. In this talk, we study the performance of two-machine lines with multiple product types, unreliable Bernoulli machines, and finite dedicated buffers, under these three policies. Exact solutions for small scale systems are derived and iteration procedures are introduced for larger ones through decomposition. It is shown that the proposed method results in high accuracy in throughput estimation. 4 - A MILP Model for Job and Labor Scheduling to Minimize Weighted Tardiness and Labor Costs for Parallel Machines Atsushi Kurimoto, Osaka University, Osaka, Japan, atsushi.kurimoto@ist.osaka-u.ac.jp, Hiroshi Morita Scheduling problems often requires optimization of multiple resources. In this paper, a two stage mixed integer linear programming (MILP) model is formulated to optimize two resources sequentially based on their priorities: machines (first level resource) and labor (second level resource). Minimizing tardiness (continuous time model) for machines and the cost of temporary employees (discrete time model) for labor are set as objectives. With the resulting schedule for machine run and individual labor assignments, the proposed model enables organization to improve their production and labor planning and assist in efficient capacity utilization. 5 - Solution Approaches for an Integrated Process Configuration, Lot Sizing and Scheduling Problem in the Packaging Industry Karim Perez Martinez, Federal University of São Carlos, Rua José Duarte de Souza, 36, São Carlos, 13564030, Brazil, karim@dep.ufscar.br, Yossiri Adulyasak, Raf Jans, Reinaldo Morabito, Eli Vitor Toso We study an integrated process configuration, lot sizing and scheduling problem in a real environment of the packaging industry. Main decisions include determining which process configurations are used, how long they are used, and in what sequence. Non-linear mathematical models are presented and symmetry- breaking constraints are applied to the formulations. A Branch-and-cut algorithm and a MIP-based heuristic are proposed to solve the problem. Results show that the exact algorithm outperforms the linearized models solved by a commercial solver for all the instances tested and the heuristic finds good quality solutions for large instances in shorter computational times. Weihang Zhu, Associate Professor, Lamar University, PO Box 10032, Beaumont, TX, 77710, United States, weihang.zhu@lamar.edu, Sujay Mahale, Ezra Wari, Alberto Marquez

1 - Sustainability Trends in Retail Sector : A Text Mining Approach Youqin Pan, Salem State University, Bertolon School of Business, 352 Lafayette Street, Salem, MA, 01970, United States, ypan@salemstate.edu This paper applies a text mining approach to discover sustainability trends in the retail sector. The findings demonstrate that green store operations, green transportation, consumer education, and employee engagement are considered as important sustainability issues or concerns in retail industry. Insights based on this study will help retail companies identify improvement opportunities by comparing its own sustainability efforts with the overall trends in the retail sector. 2 - Service Dependency and the Role of Labor in Retailing Chien-Ming Chen, Associate Professor, Nanyang Technological University-Singapore, 50 Nanyang Avenue, Singapore, 639798, Singapore, cmchen@ntu.edu.sg, Howard Hao Chun Chuang We empirically examine how service dependency of the shopper population may affect labor’s effect on retail store performance. The results show that labor’s marginal contribution to conversion and average sales are highly contingent on service dependency. Our study is based on multi-year panel data from a major retail chain in North America. 3 - Ship-from-store Operations in Omni-channel Retailing Armagan Bayram, Assistant Professor, University of Michigan Dearborn, 31035 Morlock Street, Apt 716, Livonia, MI, 48152, United States, armagan@umich.edu, Bahriye Cesaret We consider an omni-channel retailer having both online and store operations and implementing ship-from-store fulfillment model. In this model, store orders are fulfilled from store inventories, but for an online order the retailer can ship it either from the online distribution center or from any store location that maximizes its overall profit. We investigate dynamic fulfillment decisions: from which location to fulfill an online order when it arrives. We incorporate the uncertainty both in demand and in the cost of shipment to individual customers. We develop a stochastic dynamic framework and present some results on optimal fulfillment strategies. 4 - A Customer Choice Model for Estimating the Halo Effect Reza Yousefi Maragheh, University of Illinois, Urbana-Champaign, According to Halo effect, customers value the same product differently if the offered assortment varies. A customer choice model is developed that captures the pairwise effect of products on each other’s intrinsic values, i.e. the latter is a function of the offer set. This allows the model to violate the Independence of Irrelevant Alternatives property. A procedure is proposed for parameter estimation and sufficient identifiability conditions are derived. For special cases, closed form expressions are obtained. Simulation results confirm the theoretical findings. When testing the proposed model on synthetic data, it outperforms the MNL model in terms of AIC and in most cases in terms of BIC score. 5 - Environment Analysis for Global Operations Strategy in Retailing Industry Yeming Gong, Em Lyon Business School, Building B, Office 1018, 23 Avenue Guy De Collongue, Ecully, 69134, France, gong@em-lyon.com, Jiawen Liu Considering an operational decision in distribution centralization , we examine relationships between discretionary inputs of supply chain, non-discretionary inputs of environment, and organizational performance in retailing. We develop a non-discretionary DEA model to assess the corporate performance in retailing industry. Our main contribution is: (1) to provide a new method using non- discretionary DEA integrated with econometric analysis in environment analysis for GOS, and (2) to establish a new framework to understand the role of centralized distribution and its competitive environment in retailing supply chain strategy. 6 - Image-based Inventory Distribution System for Fashion Products using Convolutional Neural Network Shin Woong Sung, KAIST, 291 Daehak-ro, Yuseong-gu, Daejeon, Korea, Republic of, sw.sung@kaist.ac.kr, Hyeonjun Sim, Hyunsuk Baek, Young Jae Jang We present the image-based inventory distribution optimization method for fashion products. Due to the product variety and short product life cycles, the distribution of the fashion product to stores is challenging. Particularly for the product lines with limited quantity, the product mix is critical in the distribution decision. The proposed distribution method first identifies the visual similarity between products using convolutional neural network. Then mathematical optimization method determines the product mix and the distribution based on Urbana, IL, United States, ysfmrgh2@illinois.edu, Alexandra Chronopoulou, James Mario Davis

457