Informs Annual Meeting 2017

WA31

INFORMS Houston – 2017

3 - Examining the Effect on Firm Value of Demand-side Supply Chain Disruptions in the Social Media Era Christopher S. Kwaramba, Virginia Polytechnic Institute and State University, Blacksburg, VA, United States, shingik@vt.edu, Quinton J.Nottingham, Qianzhou Du, Alan Gang Wang chis paper investigates the long-term stock price effects of social media content following supply chain disruptions based on a sample of 45 disruption announcements made during 2010-2016. We examine the stock price effect for a two-year event time frame beginning with one stock market year before a disruption announcement to one stock market year after. By the end of the year after the disruption announcement, the average return for a stock is about −28% but returns could be as low -47%. 4 - Learn from Context: An Advanced Learning-based Approach for Health Social Media Analytics Long Xia, Blacksburg, VA, 24060, United States, longxia1@vt.edu Abstract not available. 350D Decision Analysis Contributed Session Chair: Xuan Wang, Louisiana State University, Baton Rouge, LA, United States, xwang35@lsu.edu 1 - Evaluating Classifiers Based on Cost Range to Minimize Decision Making Cost We believe using the hit rate is not enough to evaluate the performance of the classifier and it can be misleading due to overlooking an important element (decision cost). Furthermore, ROC is the most popular graph used to evaluate classifiers. However, one of the biggest ROC shortcomings is assuming equal FP and FN costs. Therefore, our goal is to reduce the total cost of decision making by selecting the classifier that has the least total misclassification cost. Nevertheless, the misclassification cost is unknown. To overcome such hurdle, we can test the data against a range of weight scale. Thus, we use the cost range and operating threshold range to show the performance differences among classifiers. 2 - Two-state Partially Observable Markov Decision Processes with Imperfect Information Tal Ben-Zvi, Stevens Institute of Technology, 1 Castle Point on Hudson, Hoboken, NJ, 07030, United States, tbenzvi@stevens.edu, Tatyana Chernonog, Tal Avinadav A process can be in either a stable or an unstable state interchangeably. The true state is unobservable and can only be inferred from observations. Several actions are available, including consulting with experts and obtaining (partial) information about the true state of the process for a fixed fee. The objective is to maximize the expected discounted value of the total future profits. We formulate the problem as a discrete-time Partially Observable Markov Decision Process (POMDP). We show that the expected profit function is convex and strictly increasing, and that the optimal policy has control limits. Also, we show that “dominance in expectation” suffices for a control limit structure. 3 - A Kalman Filtering Technique to Reduce Nonactionable Respiratory Rate and Endtidal CO2 Alarm Signals in a Continuously Monitored General Care Floor Patient John Zaleski, Chief Analytics Officer, Bernoulli, 200 Cascade Blvd, Milford, CT, 06460, United States, jzaleski@bernoullihealth.com False positive or non-actionable alarm signals along with true alarm signals are produced by Multi-parameter monitors and other bedside medical equipment within emergency departments, intensive care and general care floors. False alarm signals in these settings are a major source of distraction and a major patient safety risk. It is hypothesized that Kalman filtering is a viable means of mitigating these false alarms. An application of the Kalman filter to measured respiratory rate and end-tidal carbon dioxide illustrates its utility as an alarm signal mitigation method on data obtained from a general care floor patient undergoing continuous monitoring for a period of more than 28 hours. 4 - Perspectives on Entropy-based Weightings for Integrating Information Huguette Tran, Professor, University of North Texas, 1307 West Highland St., Denton, TX, 76201, United States, Huguette.Tran@UNT.edu, Robert Pavur In Data Fusion applications, data are often weighted according to their source reliability. A proposed weighting technique in the literature uses Shannon’s entropy. This research provides a numerical illustration to use an alternative entropy-based method based on Pal and Pal (1991)’s formulation. The alternative formulation has a bounded entropy and may provide a weighting scheme more representative of the sources’ reliability. The method is illustrated with an Ramy Alsaffar, Graduate Student in LSU, LSU, Edward Gay Apartments, 275 W. Roosevelt Street, Baton Rouge, LA, 70802, United States, ramyalsaffar@yahoo.com WA28

application of the newsvendor problem in which accurate parameter estimation is necessary for optimal profits. 5 - Comparison of Causal Inference Methods Xuan Wang, PhD Student, Louisiana State University, 2200 Business Education Complex, Nicholson Extension, Baton Rouge, LA, 70803, United States, xwang35@lsu.edu, Helmut Schneider As “Big Data” motivates an era of transformation, it not only offers many benefits, but also introduces challenges. Therefore, causal inference offers opportunities for maximizing the value of applying Big Data. This study focuses on evaluating various causal inference methods and provides recommendations. Propensity score matching, standardization, IP weighting, Bayesian network and orthogonal arrays will be discussed. 350E Manufacturing Contributed Session Chair: Brett Shields, The University of Tennessee at Knoxville, Tullahoma, TN, United States, bshields@utsi.edu 1 - Concurrent Fused Filament Fabrication with Multiple Extruders Yu Jin, University of Arkansas, 4207 Bell Engineering Center, Fayetteville, AR, 72701, United States, yujin@uark.edu, Harry Pierson, Haitao Liao Additive manufacturing (AM) is well-known for its capability in producing complex geometries. However, the fabrication speed of AM technology is considerably slower than traditional manufacturing processes A promising solution is the parallel processing of individual layers. In this research, a concurrent manufacturing strategy is proposed for improving the production speed of the Fused Filament Fabrication (FFF) process. 2 - Project Sustainability Strategies a Case of Textile Production Rhythm Wadhwa, Prof, NTNU, Teknologiveien 22, Gjoevik, 2821, Norway, rhythmwadhwa@hotmail.com A systematic literature review has been carried out focusing on cradle-to-grave textile production project sustainability. The literature review covers research published in leading journals in project management and textile production prior to 2017. Distinct perspectives have been proposed in sustianability research from the analysis.The findings of the review are complemented with an illustrative empirical case. 3 - Using Design of Experiments to Study the Significance of Different Control Parameters in Additive Manufacturing Brett A.Shields, The University of Tennessee at Knoxville, 411 B.H. Goethert Parkway, Tullahoma, TN, United States, bshields@utsi.edu, Andrew Junfang Yu In this research, the effects of control parameters on the resistivity and temperature gradient uniformity are studied for 3d printed graphene PLA. A thermal camera is used to observe the production process as the printer settings vary. Four factors (printer control parameters) are considered in the experimental design: build plate temperature, extruder temperature, print speed, and layer thickness. 351A Finance Theory and Empirics Contributed Session Chair: Hiroshi Ishijima, Chuo University, Tokyo, Japan, hiroshi.ishijima.jp@gmail.com 1 - Application of the Transmutation Operators to Parabolic Free Boundary Problems: A Numerical Method The numerical method for solving parabolic free boundary problems will be presented. The method is based on recent results from transmutation operators theory allowing one to construct efficiently a complete system of solutions for the parabolic equation, generalizing the system of heat polynomials. This techniques can be directly applied to path dependent options. The talk is based on the joint works with V. V. Kravchenko, S. Torba and J. C. Dias. “Pricing double barrier options on homogeneous diffusions: a Neumann series of Bessel functions representation” and “Solution of parabolic free boundary problems using transmuted heat polynomials” WA29 WA31 Igor Kravchenko, ISCTE-IUL, Edifício II, Av. Prof. Aníbal Bettencourt, Lisbon, 1600-189, Portugal, ivkoh@iscte.pt

451