2016 INFORMS Annual Meeting Program

MB68

INFORMS Nashville – 2016

MB69 Old Hickory- Omni Military Operations Research II Sponsored: Military Applications Sponsored Session Chair: Natalie M Scala, Towson University, 8000 York Road, Towson, Natalie M Scala, Assistant Professor, Towson University, 8000 York Road, Towson, MD, 21252, United States, nscala@towson.edu, Paul L Goethals This research applies decision analysis perspectives to cybersecurity and creates a value model for performance metrics and best practices that is supported by industry data and interviews with subject matter experts. The utility-theory based value model will include attributes and values, score metrics on their contribution to value, and provide a rank ordered list of important metrics and best practices for implementation. We illustrate the value model but contribute an overall framework that can be customized for any organization. Results will enable organizations to assess the performance of cyber systems. 2 - Efficient Benchmarking Tool Regarding Optimal Detection Of Critical Components In A Network Many mathematical and heuristic approaches have been provided to assess critical components of the network based on the network connectivity metric. Since examined objectives through this metric (e.g. minimum connectivity) have important values in many areas (e.g. immunization), proposing an effective solution framework to determine optimal values of such objectives is crucial. In this regard, we provide efficient mathematical models along with new valid inequality constraints to further decrease computational complexity compare to the most recent best models. With this improvement, we broaden the application scope of the exact solution method for the determination of critical component. 3 - OMEGA: Evaluating Effectiveness Of Proposed Systems Using Bayesian Networks Freeman Marvin, Innovative Decisions, 5848 Hunton Wood Drive, Broad Run, VA, 20137, United States, ffmarvin@innovativedecisions.com, Amanda Hepler OMEGA is a new approach for designing affordable systems architectures that meet user needs. OMEGA uses a Bayesian network of probability distributions that describes functional needs, system capabilities and customer satisfaction. Measures of Effectiveness (MOE) are combined to estimate the probability that a proposed system will meet mission needs. Additionally, OMEGA can “back cast” the system requirements necessary to achieve alternative levels of mission effectiveness. This innovative approach was developed by a collaborative team of requirements engineers and decision analysts. OMEGA is a flexible, low cost approach for conducting architecture trades and developing requirements for any kind of system. 4 - Designing An Objective Metric For Evaluating Army Unit Readiness Paul Goethals, United States Military Academy, West Point, NY, United States, paul.goethals@usma.edu, Natalie M Scala Perhaps one of the most difficult assessments to make with some level of accuracy is military readiness - it is a frequent topic of interest in defense news both in times of combat and peace. This research proposes a readiness index tailored to objectively evaluate units based upon their current status and future mission, using quality engineering tools as a foundation for measurement. A simulated comparison of the current and proposed readiness indices is provided to illustrate their differences in assessing Army units. Gokhan Karakose, University of Missouri, gkz7c@mail.missouri.edu, Ronald McGarvey MD, 21252, United States, nscala@towson.edu 1 - A Value Model For Cybersecurity Metrics

2 - Degradation Prediction Of Printed Images Ziyi Wang, Rutgers University, Piscataway, NJ, 08854, United States, ziyiwangcumtb@gmail.com, Elsayed A. Elsayed Today, a great number of images are produced by digital color printers, especially inkjet printers. Many factors lead to the degradation of such images and accurate prediction modeling of the degradation is of interest. Previous research that addresses image degradation usually measures the density loss or color change of the prints. In this presentation, the area coverage of the Neugebauer primaries for the basic four-colors (CMYK) ink-set is estimated from the spectral information of the print. A degradation model is developed to predict the area coverage loss over time. A numerical example is used to illustrate the proposed approach. 3 - Modeling Spatio-temporal Degradation Data Xiao Liu, IBM T.J. Watson Research Center, liuxiaodnn_1@hotmail.com This talk presents a modeling approach for an important type of degradation data, i.e., the degradation data collected over time and from a spatial domain. The connection between the proposed model and traditional pure time-dependent univariate stochastic degradation models is discussed, and an application example is provided. MB68 Mockingbird 4 - Omni Joint Session QSR/DM: Data analytics for system improvement I Sponsored: Quality, Statistics and Reliability/Data Mining Sponsored Session Chair: Kaibo Liu, University of Wisconsin-Madison, WI, kliu8@wisc.edu Co-chair: Haitao Liao, University of Arkansas, Fayetteville, AR, liao@uark.edu 1 - Kernel Fisher Discriminant Analysis For Uncertain Data Objects Uncertain data problems have features represented by multiple observations or their fitted PDFs. We propose measures of scatter for uncertain data objects which include covariance matrix along with within and between scatter matrices. We also propose Fisher linear discriminant and kernel Fisher discriminant for classifying uncertain data objects. 2 - An Efficient Statistical Quality Control Scheme For High- dimensional Process Sangahn Kim1, Rutgers University, Piscatawy, Piscataway, NJ, sk1389@scarletmail.rutgers.edu, Myong K Jeong, Elsayed A. Elsayed As the number of quality characteristics to be monitored increases in those complex processes, the simultaneous monitoring becomes less sensitive to the out-of-control signals especially when only a few variables are responsible for abnormal situation. We introduce a new process control chart for monitoring high-dimensional processes based on the ridge penalizing likelihood. The accurate probability distributions under null and alternative hypotheses are obtained. In addition, we find out several theoretical properties of the proposed method, and finally demonstrate the proposed chart performs well in monitoring high dimensional processes. 3 - A Nonparametric Adaptive Sampling Strategy For Online Monitoring Of Big Data Streams Xiaochen Xian, UW-Madison, Madison, WI, xxian@wisc.edu, Andi Wang, Kaibo Liu Modern and rapid advancement in sensor technology generates huge amount of data, posing unique challenges for Statistical Process Control. We propose a Nonparametric Adaptive Sampling (NAS) strategy to online monitor non-normal big data streams in the context of limited resources, such that only partial observations are available. In particular, this proposed method integrates a rank- based CUSUM scheme that corrects with the anti-rank statistics due to partial observations, which can effectively detect a wide range of possible mean shifts in all directions when each data stream follows arbitrary distribution. Two theoretical properties of the NAS algorithm are investigated. Behnam Tavakkol, Rutgers University, Piscataway, NJ, btavakkol66@gmail.com, Myong K Jeong, Susan Albin

172