Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WD67

objective of this study was to identify reproducible clusters of mild TBI patients based on rich data available at the time of the initial post-TBI patient evaluation. A sparse hierarchical clustering is applied to simultaneously select informative variables and identify underlying clusters with selected variables. Two independent datasets were utilized. Clusters found in one dataset are tried to be reproduced in another dataset. Reproducible clusters with different patient outcomes could be used to guide mTBI patient prognosis. 2 - Recent Advances in Calibration of Computer Models Rui Tuo, Atlanta, GA, 30318, United States In this talk I will show some recent advances in calibration for computer models and an application example. The goal of calibration is to identify the model parameters in deterministic computer experiments, which cannot be measured or are not available in physical experiments. In a study of the prevailing Bayesian method we find that this method may render unreasonable estimation for the calibration parameters. Inspired by a new advance in Gaussian process modeling, called orthogonal Gaussian process models, I have proposed a novel methodology for calibration. This new method is proven to enjoy nice properties. 3 - Modeling and Change Detection for Count-weighted Multi-layer Networks Hang Dong, Tsinghua University, Beijing, 100084, China, Nan Chen, Kaibo Wang In a typical network with a set of individuals, it is common to have multiple types of interactions between two individuals. In practice, these interactions are usually sparse and correlated, which is not sufficiently accounted for in the literature. This work proposes a multi-layer weighted stochastic block model (MZIP-SBM) to characterize the sparse and correlated interactions of individuals among different layers. A variational-EM algorithm is developed in order to estimate the parameters in this model. We further propose a monitoring statistic based on the score test of model parameters for change detection in multi-layer networks and evaluate the performance of this method. 4 - An Effective and Efficient Algorithm for Moving Targets Detection and Tracking with a Moving Camera Yinwei Zhang, University of Arizona, Tucson, AZ, United States, Young-Jun Son, Jian Liu Detecting and tracking moving targets with video surveillance systems is challenging, especially when using a moving camera. Conventional algorithms using projective transformation and frame differencing are inaccurate and slow. A new algorithm that combines a type of optical flow and color features is proposed to improve both detection accuracy in and tracking speed. Case studies of a variety of complex scenarios were conducted to demonstrate its effectiveness and efficiency. Joint Session QSR/Practice Curated: Blockchain Study and Application Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Hong Wan, Purdue University, West Lafayette, IN, 47907, United States 1 - A Mechanism Design Approach to Blockchain Protocols Hong Wan, Purdue University, West Lafayette, IN, 47907, United States, Abhishek Ray, Mario Ventresca, Xinqi Gao Blockchain-based systems such as cryptocurrencies are achieving widespread usage, with a market capitalization of $150B as of September 2017. However, the most prominent platforms that account for over 70% of this market - Bitcoin & Ethereum - are exhibiting increasingly lower levels of decentralization. This poses the problem of concentrating levers of consensus to a select group of agents in the system. Essentially, centralization in a system designed to be decentralized can lead to security threats such as 51% attack. In this paper, we demonstrate the use of game theory and mechanism design to find ways of solving the problem of centralization in blockchain systems. Since decentralization vs centralization is related to non-cooperative vs. cooperative behavior, using simple models based on non-cooperative and cooperative game theory we propose a way of designing payoffs in order to disincentivize certain exhibited behaviors and incentivize desired behaviors of miners or validators in such systems. n WD69 West Bldg 106A

n WD67 West Bldg 105B Aggregation of Preference Data Sponsored: Information Systems Sponsored Session Chair: Adolfo Raphael Escobedo, Arizona State University, Tempe, AZ, 85287-8809, United States 1 - A New Integer Programming Formulation and Refined Social Choice Property for Expediting the Solution to the Consensus Ranking Problem Yeawon Yoo, Arizona State University, Tempe, AZ, United States, Adolfo Raphael Escobedo We introduce an integer programming formulation for ranking aggregation with ties and compare it to a modified version of a recently developed formulation. This new formulation provides computational advantages when solving large size problems. Moreover, we develop a refined social-choice related property for group decision-making, called the Generalized Condorcet Criterion, which can be regarded as a natural extension of the Condorcet criterion. Unlike its parent properties, it is adequate for complete rankings with ties as well as for incomplete rankings. It allows us to simplify the solution process for certain types of very large instances of the NP-hard ranking aggregation problem. 2 - Rank Aggregation New Bounds for MCx Daniel Freund, Cornell University, 109 Lake St, Ithaca, NY, 14850, United States, David P. Williamson Rank aggregation is a fundamental problem in combinatorial optimization: for a set of input rankings the goal is to find a ranking that minimizes the average Kendall tau distance to the input rankings. Proposed by Dwork et al. for metasearch engines, its applications today range far beyond the original aim to problems in ML, recommendation systems and more. Though several algorithms, including a PTAS, have been proposed for these problems, it is known that some Markov chain based algorithms perform extremely well in practice. In this talk we, somewhat surprisingly, prove supra-constant lower bounds on approximation guarantees for all of them. 3 - Joint Aggregation of Cardinal and Ordinal Evaluations Adolfo Raphael Escobedo, Arizona State University, P.O. Box 878809, Tempe, AZ, 85287-8809, United States, Dorit Simona Hochbaum, Erick Moreno-Centeno An important problem in decision theory concerns the aggregation of individual cardinal and ordinal evaluations (ratings and rankings). We illustrate a new aggregation framework in the context of the 2007 MSOM’s student paper competition, which posed two unique challenges: (1) each paper was reviewed by relatively few judges, heightening the potential influence of subjective scales; (2) each judge submitted both cardinal and ordinal evaluations of their assigned papers. To address these challenges, we employ axiomatic distances and derive new non-convex exact formulations and convex relaxations. The solution methodologies are further tested and analyzed via randomly generated instances. 4 - Elaborating or Aggregating? The Joint Effects of Group Decision-making Structure and Systematic Errors on the Value of Group Interactions Wenjie Tang, NUS Business School, 15 Kent Ridge Dr, Singapore, 119245, Singapore, Steffen Keck We explore when group interactions will have a positive effect on the accuracy of quantitative judgments. The results from two laboratory experiments revealed that compared to a statistical aggregation of individual judgments, group interactions only provide values when the level of systematic error is high and when there is a designated group leader. Moreover, our analysis showed that this effect was mediated by information elaborationùwhich was generally higher in leader groups but only had a significant effect on the value of group interactions when there was a high level of systematic error among group members, and not otherwise. Joint Session QSR/Practice Curated: Data Analytics in Complex Systems: Methodology and Applications Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Changyue Song, University of Wisconsin-Madison, Madison, WI, 53715, United States 1 - A Cross-study Analysis for Reproducible Sub-classification of Mild Traumatic Brain Injury Bing Si, Arizona State University, Mesa, AZ, 85201, United States The current stratification of traumatic brain injury (TBI) into “mildö, “moderateö, or “severe does not adequately account for the patient heterogeneity. The n WD68 West Bldg 105C

516