Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WA22

3 - Generative Adversarial Network Formulations with Convergence Guarantees Maziar Sanjabi, USC, Los Angeles, CA, United States

5 - Multinational Merger and Acquisition and Market Value of China’s Service Enterprises in the Belt and Road Strip Hao Lu, PhD Student, University of Science and Technology of China, 96 JinZhai Road, Baohe District, Hefei, 230026, China The Belt and Road strip (BRS) provides a broad stage for Chinese service enterprises to enter into M&As overseas. However, whether such multinational M&As achieved the expected market value return? This paper uses event study methods to analyse the development of multinational M&As in the service industry in BRS for the period 2013-2017. We found that since 2013, most of China’s multinational service M&As have achieved significantly positive abnormal returns (ARs). The ARs of M&As in the technology services and public services were found to be more significant than other sub-industries of service. Among the factors that affect AR, the ratio of M&A contract size is the most significantly positive.

Generative Adversarial Networks (GANs) are one of the most practical strategies to learn data distributions. A popular GAN formulation is to use Wasserstein distance as a metric between probability distributions. Unfortunately, minimizing this distance is difficult as its objective is non-convex and non-smooth. In this work, we use smooth approximations of this objective. We show that obtaining gradient information of the smoothed formulations is easy. Based on this observation, we propose a class of first-order algorithms with guaranteed theoretical convergence to stationarity. We apply our method to MNIST and CIFAR-10 datasets and show its effectiveness. n WA22 North Bldg 130 Practice- Pricing and Revenue Management I Contributed Session Chair: Hao Lu, University of Science and Technology of China, 96 JinZhai Road, Baohe District, Hefei, 230026, China 1 - Which Outlier Detection Approach is Better for Estimating the Elasticity Value Shuguang Ji, Sr. Data Scientist, Delta Airlines, Atlanta, GA, 30339, United States In cases where outliers are present, finding an accurate model to make prediction becomes more difficult. In practice, the mean +/- 3 standard deviations remains common method for outlier detection. Unfortunately, 3 problems can be identified when using this method. Firstly, it assumes that the distribution is normal. Secondly, the mean and standard deviation are strongly impacted by outliers. Thirdly, this method is very unlikely to detect outliers in small samples. In this presentation, the author wants to review the popular outlier detection methods for the elasticity estimation at first. Then the most effective detection Yingda Zhai, The University of Texas at Austin, Austin, TX, 78703, United States, Maxwell B. Stinchcombe, Andrew B. Whinston Motivated by unprecedented surging demand for computing resource, we consider a monopoly provider with queue technology prices and allocates computing resource in a complex environment. We establish a service model with no aggregate risk which allows provider accurately to predict and manage queue capacity. A profit-maximizing provider throttles service by (i) limiting overall market supply and (ii) tranching her capacity in terms of service quantity and interruption. We then propose an appealing auction mechanism in the cloud computing spot market to implement the profit-maximizing allocation where users find it a weakly dominant bidding strategy to bid the optimal prices. 3 - Dynamic Pricing Strategy and Simulation Research Considering Strategic Consumers in Perfect Competition Electricity Market Lingchunzi Li, Huazhong University of Science and Technology, School of Management, No. 1037, Luoyu Road, Hongshan District, Hubei Province, China, Wuhan, 430074, China, Haijun Wang, Nancy Shuojia Guo, Penghong Cheng China is in the key stage of electricity market reform. This paper is to explore an appropriate pricing strategy to provide reference for reformation and power pricing research in China. Under centralized transactions, a game model considering customers’ strategic choice behavior is established. And a time- sharing dynamic pricing scheme is solved to maximize both power enterprises’ equilibrium revenue. Simulation results show that the competition between power producers can reduce power price and increase power supply volume effectively. The model we built and the factor of equilibrium validity we introduced are innovations in this paper. 4 - An Optimization Framework for Pricing of Port Services: Case of Major Ports in India Deepankar Sinha, Associate Professor, Indian Institute of Foreign Trade, 1583 Madurdah, Kolkata, 700107, India Ports enable inter and intra-modal cargo transshipment to and from ships. The cost incurred in ports impact the total landed cost of any maritime cargo. The ports need to achieve three basic objectives. One, it needs to be efficient in achieving the desired level of productivity with optimal use of resources; second, minimize cost of operation; and third minimize total landed cost of shippers. So far, studies and researches have focused on part of these objectives. In this paper an attempt has been made to encompass all the three objectives. It proposes to use a non-parametric approach to develop a multi-objective optimization model to arrive at a rational pricing framework for port services. methods will be recommended according to the cases in practice. 2 - Pricing and Service Throttling in Cloud Computing

n WA23 North Bldg 131A Systemic Risk and Financial Risk Management

Sponsored: Finance Sponsored Session

Chair: Luitgard Veraart, London School of Economics and Political Science, Houghton Street, London, WC2A 2AE, United Kingdom 1 - Optimal Portfolio Allocations in a Heterogeneous Banking System Marko Weber, Columbia University, New York, NY, United States We study the portfolio choice implications of leverage constrained banks, which may need to deleverage in response to price shocks to satisfy regulatory requirements. Banks select their asset holdings in order to minimize their expected execution costs. Consistent with the classic theory of portfolio selection, diversification is optimal if each bank neglects the impact caused by the other agents’ liquidation actions. If banks are heterogeneous in their leverage ratios, in equilibrium they reduce portfolio overlapping and seek diversity, at the expenses of sacrificing diversification benefits on the individual level. The bank’s equilibrium allocation is not socially efficient. A benevolent social planner aiming at minimizing liquidation costs should provide banks with incentives to increase their diversity. 2 - Interbank Clearing in Financial Networks with Multiple Maturities Luitgard Veraart, Associate Professor, London School of Economics We consider the problem of systemic risk assessment in interbank networks in which interbank liabilities can have multiple maturities. We develop a clearing mechanism for the interbank liabilities to deal with the default of one or more market participants. Our approach generalises the clearing approach for the single maturity setting by Eisenberg & Noe (2001). We show that in the context of multiple maturities, specifying a set of defaulted banks is challenging. We propose two approaches to overcome this challenge: an algorithmic and a functional approach. Our analysis permits construction of simple dynamic clearing models. 3 - Buffered Probability of Exceedance (bPOE) Ratings Stan Uryasev, University of Florida-Gainesville, AOD, 303 Weilhall, Gainesville, FL, 32611, United States, Giorgi Pertaia Credit ratings are widely used by investors to assess the credit risk of a security. The financial crisis of 2008 showed that credit ratings might not measure the risk appropriately for the synthetic instruments such as CDOs. This paper presents a new rating assignment model based on Buffered Probability of Exceedance, that is an improvement over the current POE based methodology. 4 - Risk in Production Networks Peng-Chu Chen, The University of Hong Kong, Kowloon, Hong Kong We investigate systemic risk in a production network, where firms are connected through demand and supply orders. A financial shock causing disruptions to some firms propagates through the supply-chain network via the input-output linkages between suppliers and customers. Each firm in the network takes contingency plans to mitigate the impact generated from disruptions and reroutes orders through different suppliers. An equilibrium is reached when contagion from disruption stops. We develop an algorithm to recover the equilibrium with the greatest amount of delivered orders. We then analyze the impact of industrial integration on systemic risk in the production network.” and Political Science, Houghton Street, Department of Mathematics, London, WC2A2AE, United Kingdom, Michael Kusnetsov

418