Informs Annual Meeting 2017

WB57

INFORMS Houston – 2017

2 - A Complete Asset Market Approach to the Determination of Equilibria for Incomplete Asset Markets Chuangyin Dang, City University of Hong Kong, Dept of Systems Eng & Eng Mgmt, 83 Tat Chee Avenue, Kowloon, Hong Kong, mecdang@cityu.edu.hk In incomplete asset markets, there are an insufficient number of contingent claims instruments to provide households with all potentially desirable credit arrangements. This insufficiency can cause discontinuity of the excess demand function at prices for which the asset return matrix loses its full rank. To overcome this discontinuity, we make use of the product of the asset return matrix and its transpose as a new asset return matrix. With the introduction of an extra variable and a convex combination of the new asset return matrix and an identity matrix, we establish an artificial complete asset market, which assures the existence of a smooth path to an economic equilibrium. 3 - E-commerce Platforms and International Trade: A Large-scale Field Experiment Xiang Hui, MIT, Cambridge, MA, 02139, United States, xianghui@mit.edu Information technology has enabled businesses of all sizes to sell globally. Exploiting a site-wide randomized experiment on eBay, I show that an e- commerce platform could further increase on-site international trade by 2.9% in the short run and 12.3% in the long run through integrating an existing service of handling customs clearance and international shipping. The increase comes from “disadvantaged” groups (small and medium sellers, distant countries, and less profitable products). Foreign consumers benefit from an 8.3% larger product variety. I discuss the role of “nudging” and reducing export entry costs for facilitating exports from small firms. 4 - Generalized Additive Model with Embedded Variable Selection for Bankruptcy Prediction Laura María García, University De Los Andes, Bogota, Colombia, garcia.lm@uniandes.edu.co, Carlos Felipe Valencia Arboleda, Sergio Andrés Cabrales In this study, we explore the trade-off between interpretability and predictive performance in the problem of bankruptcy prediction using finantial ratios. We propose a generalized additive model with an innovative embedded variable selection methodology that selects zero, linear or non-linear effects through penalization. Our experimental results demonstrate that the proposed model performs well in terms of predictive power and generalization, providing a good balance between flexibility and interpretability. The findings of this research show that non-linear methods are necessary to understand the behavior of the retail industry in Colombia. 5 - Cost-efficiency of Brazilian Water Distribution and Sewage Collection Services: Some Recent Evidence Fernando Garcia Freitas, Advisor, National Confederation of Services, Saint Hilaire, 118 - 22, Sao Paulo, 01423040, Brazil, fernando.garcia.freitas@gmail.com This paper discusses the determinants of production costs in residential water distribution and sewage collection in Brazil. We estimate two distinct stochastic cost frontier models Aigner, Lovell and Schmidt (1977) and Battese and Coelli (1995) for a sample of 23,329 observations from 2009 to 2014. A translog specification is adopted for both models. Cost depends on the relative price of inputs and the level of services provided. Estimations reveal statistically significant parameters for most of variables in the two distinct models and that inefficiency is heterogeneous among groups of firms. The main result obtained is that open capital firms are more efficient than closed capital ones.

2 - Scheduling Software Updates for Connected Cars with Limited Availability Carlos E.De Andrade, AT&T Labs Research, 200 Laurel Avenue South, A5-1E33, Middletwon, NJ, 07748, United States, cea@research.att.com, Simon D.Byers, Vijay Gopalakrishnan, Emir Halepovic, David J. Poole, Lien Tran, Chris Volinsky The number of connected cars has increased over the past years, and they became an important component of Internet-of-Things. Such vehicles use the cellular networks for several activities, and among them, the Firmware Over-The-Air updates could be potentially challenging to the network. With millions of connected cars expected to be deployed over the next years, it is important to understand their impact on cellular networks, as much as create schedules for such downloads. We present a new scheduling model that accommodates the constraints for such scenario, which usually does not appear in other scheduling problems in the literature, and comment the peculiarities of such solution. 3 - Reallocatable Multi-skilled Resource Constrained Project Scheduling Problem: Mathematical Modeling and Solution Approaches ZhenTao Hu, HUST, Wuhan, China, gentlehzt@163.com In this paper we propose an extension of the multi-skilled resource constrained project scheduling problem(MSRCPSP): the Reallocatable multi-skilled resource constrained project scheduling problem(RMSRCPSP). In the RMSRCPSP the resources in use are allowed to be replaced by other resources that master skills the processing activities need. To solve this problem, we apply a priority-based heuristic algorithm, which makes use of resource weight and extend the parallel scheme by introducing a resource shifting procedure. A series of computational tests including small scale and large scale instances shows that the heuristics algorithm is very effective. 362E Facility Location II Sponsored: TSL, Facility Logistics Sponsored Session Chair: Tan C Miller, Rider University, 12 Winding Way, Morris Plains, NJ, 07950, United States, tanjean@verizon.net 1 - The Mean-var Median Problem on a Network with Probabilistic Demand Weights Jiamin Wang, Professor, Long Island University, Post Campus, Roth Hall 202, 720 Northern Boulevard, Brookville, NY, 11548-1300, United States, jiamin.wang@liu.edu, Chunlin Xin We consider a network facility location problem where the number of potential customers at each node is a random variable. The classical median problem is extended to minimize the expected network length subject to a value-at-risk constraint. Different risk measures are compared to motivate the study. We analyze the property of the optimal solution and identify a finite set of dominant points. Efficient solution procedures are then developed to solve the model. The solutions to the problem are also examined to get managerial insights. 2 - A Cutting Plane Method for Competitive Facility Location under Random Utility Maximization Models Tien Anh Mai, Postdoctoral researcher, CP 6128 Succursale Centre-Ville, Apt 8, Montreal, QC, H3W1C5, Canada, maitien86@gmail.com This work concerns the facility location problem in competitive market, where the demand of users is captured by a random utility maximization model. Existing approaches often use the logit model, and formulate the problem in a mixed- integer linear programming model. In this paper, we formulate the problem under the mixed logit model, which is more general and fully flexible. We propose a new approach based on a cutting plane method to exactly and quickly solve the nonlinear problem. We test our algorithm and show that our algorithm is much faster, compared to existing approaches. 3 - Optimizing Product Allocation in a Milkrun Picking System Jelmer Pier van der Gaast, Postdoctoral Researcher, Rijksuniversiteit Groningen, Nettelbosje 2, Groningen, 9747 AE, Netherlands, j.p.van.der.gaast@rug.nl In a milkrun order picking system, an order picker picks orders that arrive in real time during the picking process; by dynamically changing the stops on the picker’s current picking route. The advantage of milkrun picking is that it reduces order picking set-up time and worker travel time compared to conventional systems. We study order throughput times of multi-line orders and model the system as a cyclic polling system with simultaneous batch arrivals. This allows us to study the effect of different product allocations in an optimization framework for a real-world application. WB58

WB57

362D Scheduling Contributed Session

Chair: ZhenTao Hu, HUST, Wuhan, China, gentlehzt@163.com 1 - Multiple Objective Optimization for Bus Scheduling Problem in Vietnam Ho Thanh Phong, International University, Ho Chi Minh City, Vietnam, KP6, Linh Trung, Thu Duc Dist., Ho Chi Minh City, Viet Nam, htphong@hcmiu.edu.vn, Nguyen Thanh Phong The multi-objective transit network design and frequency setting problem (TNDFSP) involves finding a set of routes constituting a public transit network with three objectives, namely, minimizing user cost, minimizing operator cost and minimizing the congestion score. A Genetic Algorithm (GA) was applied to solve that multi-objective problem, resulting the selection of the “best compromise” solution among the non-dominated transit network solutions. The research was applied and expected to improve the bus service in Ho Chi Minh city, Vietnam.

494