Informs Annual Meeting 2017

MD70

INFORMS Houston – 2017

MD68

shortage mitigation while adhering to constraints that force the physical and topographical structures of the river. Solutions combine upstream pumping and new storage in existing reservoirs, as well as storage in new surface and subsurface reservoirs, and provide prescriptive, rather than descriptive, information. 2 - Coordinated Long-term Planning of Interdependent Energy and Water Systems Shanshan Hou, University Of Arizona, 1670 N.Forgeus Avenue, Tucson, AZ, 85716, United States, shanshanh@email.arizona.edu, Neng Fan Water is an important player for decisions in energy system, and energy network supply power for water distribution and extraction. These two infrastructure systems could be modeled as interdependent network. We consider both their operations, and planning period, which analyzes building the possible facilities, adding generators and transmission lines. As long-term planning and real-time operations have mutual and conflicting impacts in their decision making, we propose optimization model to simulate the coordinated system. The objective is to save money for planning and operations. We test the model in the example of IEEE 6-Bus and EPANet 1-Tank, as well as IEEE RTS-96 and EPANet 3-Tank. 3 - Data-Driven Distributionally Robust Water Allocation under Climate Uncertainty Jangho Park, The Ohio State University, 1971 Neil Avenue, Columbus, OH, 43210-1271, United States, park.1814@osu.edu, David Love, Guzin Bayraksan We apply data-driven distributionally robust stochastic optimization to a real- world water resources allocation model. We allocate Colorado River water to different users by considering possible extreme water shortages and construction of additional water treatment infrastructures until 2050. The system has two main uncertainties, future water demand, and supply. They are interconnected with other uncertain factors such as population growth and climate variability. We provide a statistical framework to predict the future uncertainties. Our data- driven decision provides a cost-benefit analysis for the government to make the long-term city plans and introduce new infrastructures. 4 - A Methodology to Assess the Energy-water Nexus in the UW- Madison, United States Roshi Nategh, PhD, Purdue, West Lafayette, IN, United States, Rnateghi@purdue.edu Despite the high level of interdependency in energy-water systems, they have been designed, managed and operated by separate entities, and researched independently. However, ensuring the reliability and resiliency of both resources commands knowledge of their interconnectedness. While many research endeavors have focused on addressing the energy-water nexus on the supply side, there is a scant body of knowledge on energy-water demand. In this talk, we present a multivariate, nonparametric technique - leveraging data from various publicly available sources such as the EIA, USGS and US Census - to assess the energy-water nexus in the U.S. 371E Data Mining Contributed Session Chair: Jilei Zhou, Guanghua School of Management, Peking University, Beijing, China, 1501110963@pku.edu.cn 1 - Sustainable Transportation Planning Mohammad Ali Asudegi, UTK, 525 John D. Tickle Building, 851 Neyland Drive, Knoxville, TN, 37996-2315, United States, masudegi@vols.utk.edu, Rapinder Sawhney A new perspective of sustainable transportation planning is introduced to help with expanding acceptance of green transportation solutions. 2 - Mining Consumer Shopping Patterns with Mobility and Social Network Data Sharon Xu, Massachusetts Institute of Technology, UW-Madison, 235 Albany Street, #3023A, Cambridge, MA, 02139, United States, sharon_x@mit.edu, Marta C.Gonzalez The majority of work in matrix factorization has focused on factorizing a single matrix, making it difficult to exploit the numerous data sources often available. In this talk, we present a collective factorization approach to predict shopping behavior through the addition of consumer mobility and network data. We model how shopping preferences change temporally, jointly factorizing the monthly user-purchase matrix (UVT), user-location matrix (UWT) and social network (UZT) with a shared user latent feature space U. Through this approach, we improve upon single matrix factorization while connecting multiple perspectives of user factors to better understand consumer lifestyles. MD70

371C Data Analytics for System Improvement I Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Xiaochen Xian, UW-Madison, Madison, WI, 53705, United States, xxian@wisc.edu Co-Chair: Kaibo Liu, UW-Madison, Madison, WI, 53706, UW-Madison, United States, jacoblkb@gmail.com 1 - Improving Food Assistance System Service by Considering Measurement Error Haomiao Yang, University of Arizona, Tucson, AZ, United States, haomiaoyang@email.arizona.edu, Daoqin Tong, Jian Liu Generalized linear regression models are usually used to model the service quality of food assistance systems. Conventional estimation methods are based on the assumption that the model covariates are free of errors. In this research, a new algorithm is proposed to explicitly consider two type of errors associated with regression covariates and improve model estimation accuracy. The case study shows the effectiveness. 2 - A Multimodality Factor Mixture Model with Hierarchically- structured Sparsity for Disease Subtype Discovery Bing Si, Arizona State University, Tempe, AZ, 85287, United States, bingsi@asu.edu, Jing Li Medical imaging technology has revolutionized health care and plays a pivotal role in Precision Medicine (PM). Disease subtype discovery paves the way toward PM by stratifying patients into subgroups with similar clinical characteristics, prognostics, or/and treatment response. To enable subtype discovery from multimodality imaging data, we propose a Multimodality Factor Mixture Model with a double-L21 penalty to achieve hierarchical selection of informative modalities and features and adopt an efficient Group-wise Majorization Descent algorithm for model estimation. Its performance is demonstrated by both simulation studies and a dataset collected by 2 U.S. medical institutions. 3 - Multifractal and Lacunarity Analysis of Image Profile from UW- Madison, Ultra-precision Machining Farhad Imani, PhD Student, Pennsylvania State University, 233, Leonhard Building, 310 S.Barnard St, University Park, PA, 16802, United States, fxi1@psu.edu, Hui Yang In this research, we developed a novel method of multifractal and lacunarity analysis which can capture the defects from the in-situ image of ultra-precision machining. Multifractal measures the variation in a layer of image and lacunarity measures heterogeneity to complement multifractal analysis and work together these statistical quantities can characterize patterns extracted from images and very precisely detect the existence of an underlying multiplicative process. The multifractal and lacunarity analysis have shown to have strong potential to be generally applicable as new process monitoring and control tool in a variety of domains such as bioengineering and nanoengineering. 4 - Online Monitoring Big Data Streams: A Rank-based UW-Madison, Sampling Algorithm Xiaochen Xian, UW-Madison, 936 Eagle Heights Apt. C, Madison, WI, 53705, United States, xxian@wisc.edu, Kaibo Liu With the rapid advancement of sensor technology, a huge amount of data is generated in modern engineering applications, which poses new and unique challenges for Statistical Process Control (SPC). In this paper, we propose an online sampling strategy to online monitor big data streams in the context of limited resources, where only partial observations are available at each acquisition time. Both simulations and case studies are conducted under different scenarios to illustrate and evaluate the performance of the proposed method. 371D Optimizing Water Resources Sponsored: Energy, Natural Res & the Environment Environment & Sustainability Sponsored Session Chair: Alexandra M Newman, Colorado School of Mines, Golden, CO, 80401, United States, anewman@mines.edu 1 - Designing River Basin Storage using Optimization Andy Burrow, MS, Colorado School of Mines, Golden, CO, 80401, United States, aburrow@mymail.mines.edu The ways in which a growing population increases hydrologic demand are often evaluated using simulation models. This research uses data produced by the State of Colorado’s Stream Simulation Model as input to an optimization model to determine the flow of unappropriated water so as to minimize the cost of water MD69

261