Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

POSTER SESSION

37 - Research on Baoji’S Titanium Industry Chain Patent Navigation Based on Social Network Analysis Method Hongru Yan, Northwestern Polytechnical University, Xi’an China, China Patent right is the industrial intellectual property right which can best reflect the innovation level of science and technology. This paper collected and retrieved the patent information of Baoji titanium industry from 2012 to 2016 and used patent centrality analysis and patentee centrality analysis to clarify the research highlight of this field and relationship between main patentees. Finally, conclusions and recommendations have been given from 4 aspects, which contains focusing on the downstream development of Baoji titanium industry chain, facilitating the high-end extension of Baoji titanium industry chain and so on . 38 - Stochastic Corrective Risk-based Optimal Power Flow Thanh To, Clemson University, Clemson, SC, 29631, United States We propose a model that quantifies and controls the total system line overflow caused by contingencies via means of Conditional Value-at-Risk (CVaR) constraint. We then incorporate the wind uncertainty into our model and perform computational experiments. 39 - A Case Study on Condition-based Maintenance through Machine Learning on Sensor Data Bram Steurtewagen, Ghent University, Steenweg Deinze 124a, Nazareth, 9810, Belgium In the petrochemical industry, we encounter a vast landscape of sensor data that is not being used to its full potential. This data, enabled by the increase in processing power and analytical frameworks over the years, can be utilized to create predictive models and perform root cause analysis where before these domains lacked statistical and numerical input. This case study aims to implement Industry 4.0 standards by leveraging analytical techniques on real sensor data for predicting machinery failures and identifying possible causes of breakdowns. The research shows that this approach can match plant operator instincts while also providing new insights and a more fine-tuned triggering process. 40 - Optimal Planning of the Joint Placement of Photovoltaic Panels and Green Roofs Under Climate Change Uncertainty Mohammad Ramshani, University of Tennessee, Knoxville, TN, 37920, United States, Anahita Khojandi, Xueping Li, Olufemi A. Omitaomu Photovoltaic (PV) panels directly convert sunlight into electricity; but, sunlight also heats the panels, negatively impacting their efficiency. Green roofs are vegetative layers grown on rooftops, mainly to provide added insulation on the roof to save energy. Green roofs also cool near-surface air temperature. Hence, the joint installation of PV panels and green roofs may potentially lead to higher efficiency of PV panels in certain climates. We develop a two-stage stochastic programming model to optimally place PV panels and green roofs under climate change uncertainty to maximize the overall profit from energy generated and saved. 41 - Integrating Multimodal Transport into Bioethanol Supply Chain Design Under Supply and Demand Uncertainties Yong Shin Park, St. Edward’s University, Austin, TX, United States, Yuan Xu A mixed integer two-stage stochastic optimization model was developed that integrates multimodal transport into the lignocellulosic-based bioethanol supply chain in an uncertain decision environment. The goal was to maximize the expected profit to support strategic and tactical decision planning. A case study based on North Dakota state in the U.S. was tested to explore the potential of multimodal transport for bioethanol supply chain. The model captures the challenges with supplying biomass to a biorefinery and shipping biofuel to demand centers under supply and demand uncertainty. Sensitivity analyses was performed to gain managerial insights on how profit changes due to uncertainties. 42 - Deviations Between the Chinese and International Gold Price: Role of Fundamentals, Contagion and Financial Shocks Zhicheng Liang, HKU, Hong Kong, Hong Kong, Junwei Wang China has been both the world’s largest gold producer and consumer since 2013. With the second largest exchange-traded market in the world, however, China’s gold price frequently diverges from international gold price. This study explores the impact of fundamentals, contagion and financial shocks on the price gap using an extended GARCH model. The empirical study shows that the exchange rate of renminbi plays an important role in moving the price gap. The global contagion factors associate more with the level than the volatility of the price gap. For financial shocks, policy shocks do not present significant influence on the price gap, but black swan events do. 43 - Robust Portfolio Optimization with Options Hedieh Ashrafi, Southern Methodist University, EMIS, PO Box 750123, Dallas, TX, 75275-0123, United States, Aurelie Thiele We consider the problem of maximizing the worst-case return of a portfolio when the manager can invest in stocks as well as European options on those stocks, and the stock returns are modeled using an uncertainty set approach. We present theoretical results regarding the structure of the optimal allocation as well as numerical results comparing our approach to several benchmarks, showing that our approach performs well in practice.

44 - A Machine Lfarning Approach to Optimize Risk Prediction and Binary Classification for Medical Diagnosis Minmin Zhang, University of Michigan, Ann Arbor, MI, 48105, United States, Zheng Zhang, Brian T. Denton We present a new continuous optimization model which maximizes the accuracy of classification and prediction respectively in different risk regions. We show that the cost function is continuously differentiable under appropriate parameters. Our proposed model offers improvements over common approaches such as regression and Lasso on the basis of true positive and true negative classification rates and calibration measured by the Brier Score in different risk regions. We illustrate our approach based on real data in the context of medical diagnosis. 45 - Improving Non-monotonicity Issues in Predicting Survival Probability of Transplant Patients Hamidreza Ahady Dolatsara, Auburn University, Auburn, AL, 36830, United States, Fadel Mounir Megahed, Ying-Ju Tessa Chen, Richard Sesek, John Evans, Ashish Gupta This study describes non-monotonic issues in predicting survival probabilities of transplant patients over a time horizon. Survival estimates of patients receiving heart transplants over a consecutive time horizon are evaluated. Various non- monotonicity issues are first investigated. Isotonic regression is adopted to resolve these issues. Prediction performance of before and after utilizing isotonic regression is finally analyzed. 45B - Competitive Pricing of Sea and Rail Transport Based on the Selection of Shippers WeiDong Dai, School of Transportation and Logistic,Southwest Jiaotong University,Chengdu 611755, Sichuan, China, Chengdu, China Railway pricing is vital to the development railway. To a certain degree,the partaking rate is determined by price. It is important to establish tariffs reasonably.This essay introduces the factors that customers will consider when choosing means of transportation, and constructs a generalized cost function. Then ,we use logit model to determine the market share rate of various means of transport, and we build a game model based on the revenue maximization of various means of transport. Finally, we use Chengdu-European Express Railway and a shipping company as a example to analyze to obtain the equilibrium freight rates and sharing rates of Chengdu-European Express Railway and shipping company. 46 - Automatic Surrogate Modeling for Complex Engineering Systems Reza Alizadeh, University of Oklahoma, Norman, OK, United States There are three major selection steps in surrogate modeling, including the surrogate type, metrics and test problems. From the critical evaluation of the literature, we identified two main gaps: 1) The multiple surrogate always composed by KRG, RSM and RBNN. No research studied whether there is any other types of multi-surrogate can perform better; 2) Although many comparison works have done, there are no ruled-based knowledge generated to guide others to select a better surrogate model for their problem. Results show that Multi- surrogates has the highest accuracy and robustness. Also, the generated rules are inputs for a knowledge-based platform to make the surrogate modeling process automated. 47 - Online Learning for Stochastic Game with Risk Preferences Wenjie Huang, National University of Singapore, Singapore, Singapore, Pham Viet Hai, William Haskell In this work, we study the stochastic games considering risk preferences of different players. We show the existence of risk-aware Markov perfect equilibrium and develop a reinforcement learning type algorithm to solve the equilibrium online. Our algorithm is based on solving stochastic saddle-point problem for risk estimation and doing Nash Q-learning for finding the equilibrium. We show that our game model covers wide class of risk measures and we derive the almost sure convergence and computational complexity result for our algorithm. 48 - Optimization of Preferred Boarding Groups by Markov Chains Viacheslav V. Kalashnikov, Professor & Researcher, Tecnologico de Monterrey, (ITESM), Campus Monterrey, Ave Eugenio Garza Sada Examine a bilevel optimization problem, in which the upper-level is an airline’s management governing the price and proportion of preferred boarding passes (PBP) available. Passengers seek the equilibrium at the lower level of the model by maximizing their payoff functions thus generating a Markov chain’s transition matrix. The steady-state probabilities of the Markov chain are used at the upper level as the powers of the pertinent classes into which the set of passengers is partitioned. The existence of solutions has been established and simple numerical experiments made. As a result, some practical recommendations concerning the optimal portions of PBPs and their price can be developed. 2501 Sur, Monterrey, Nuevo Leon, 64849, Mexico, Edgar Camacho-Esparza, Nataliya Kalashnykova, Vladik Kreinovich

180