2016 INFORMS Annual Meeting Program

SC09

INFORMS Nashville – 2016

3 - Data Mining For Result Prediction In Sports Kyuhan Lee, Seoul National University, Seoul, Korea, Republic of, kyuhanlee0119@gmail.com Jinsoo Park, Buomsoo Kim

4 - Allocating Countermeasures To Defend Water Distribution Systems Against Terrorist Attack Jacob Monroe, North Carolina State University, Raleigh, NC, United States, jgmonroe@ncsu.edu Elizabeth Ramsey, Emily Zechman Berglund An agent-based model is developed to simulate the attack and defense of a water distribution system to analyze security resource allocation strategies for protecting against chemical contamination events. A single period attacker-defender game is simulated, in which an attacker seeks to contaminate a system node, and a group of defenders seek to minimize the public health impact from attack. Terrorist decisions are simulated using a multi-attribute utility function. The utility manager assigns personnel and security equipment to nodes using one of three security resource allocation strategies. SC09 103B-MCC Energy and Environmental Policy Sponsored: Energy, Natural Res & the Environment I Environment & Sustainability Sponsored Session Chair: Yihsu Chen, University of California Santa Cruz, 1156 High Street, M/S SOE3, Santa Cruz, CA, 95064, United States, yihsuchen@ucsc.edu 1 - The Cost Of Reaching Mexicos Climate Change Goals Rodrigo Mercado Fernandez, UMass, Amherst, MA, rodmerfdez@gmail.com This paper analyzes the cost of Mexico reaching its climate change emissions goals, using integrated assessment models, and looks at how this will affect the electricity generation portfolio. These results are compared with the predicted impacts that Mexico’s current policies will have on emissions and generation. Lastly this paper identifies policy changes that could help Mexico reach its long- term emissions goals for 2030 and 2050. 2 - On The Inefficiencies Of The US Federal Clean Power Plan Duan Zhang, University of California Santa Cruz, Santa Cruz, CA, United States, dzhang33@ucsc.edu, Yihsu Chen, Makoto Tanaka The performance-based standard under the US federal Clean Power Plan relies on trading the emission rate credits (ERCs), which represent the equivalent MWh of energy generated or saved with zero associated CO2 emissions, to equating marginal abatement costs across generating units. We show theoretically the equivalence between the ERCs and the traditional mass-based trading when states are subject to their own performance-based standards. We also identify the conditions under which the inefficiency of the performance-based standard might arise, leading to a divergence of permit prices across states. A numerical 3-node model was built to illustrate our findings. 3 - Feed-in Tariffs Vs. Renewable Portfolio Standards: The Effect Of Market Power Recently policies for promoting renewable energy, e.g., feed-in tariffs (FIT) and renewable portfolio standards (RPS) have been introduced in various countries. In this work, we examine an effect of market power in the electricity market on FIT and RPS by bi-level model. For lower level, generation outputs for renewable and non-renewable generators are decided by maximizing their profits whereas for upper level, the fixed price of FIT and the RPS requirement are derived by maximizing a social welfare. In addition, we show how the number of firms affects the fixed price and the requirement. 4 - Tradable Performance-based Co2 Emissions Standards: Walking On Thin Ice? Mari Ito, Tokyo University of Science, Noda, Japan, mariito@rs.tus.ac.jp, Ryuta Takashima, Makoto Tanaka, Yihsu Chen US federal Clean Power Plan (CPP) stipulates a state-specific performance-based CO2 standard and offers considerable flexibility to the states in achieving the target. We analyze the tradable performance standards and related mass-based standard when they are subject to imperfect competition by formulating them either as a complementarity problem or a mathematical program with equilibrium constraints (MPEC). The MPEC is solved as mixed integer problems with a binary expansion. We show that while the cross-subsidy inherent in the performance-based standard that might effectively reduce power prices, it could in inflate energy demand, thereby rendering permits scarce. Yihsu Chen, University of California Santa Cruz, yihsuchen@ucsc.edu, Afzal Siddiqui, Makoto Tanaka

The expansion of sports betting has extensively contributed to the increase of public interest on sports result prediction. In academia, as statistical data respecting sports games are readily accessible, abounding research has been conducted regarding the subject. In this paper, unlike the past studies focusing on limited types of data, we use a comprehensive set of data, including statistical data as well as text data, to enhance the accuracy of sports result prediction. We expect that our prediction model produces a preferable outcome comparing to the models of previous research. 4 - Predicting Users’ Continuous Participation In Online Health Virtual Community: Demographic And Content Cues Online health community (OHC) is a platform where people with similar health conditions gather virtually to ask questions, share experiences, provide support, and exchange healthcare knowledge. To be effective, the OHCs have to maintain active continuance participation from their users. The purpose of this study is to identify factors that affect the users’ continuance participation. Specifically, we attempt to use data analytics techniques to identify the demographic and content cues that affect the users’ continuance participation. The findings of our research help community managers deploy various strategies to encourage the continuance participation of different types of members. Yanyan Shang, Dakota State University, Madison, SD, United States, yshang@pluto.dsu.edu, Jun Liu, Iljoo Kim Chair: Pavithra Harsha, IBM Research, 1101 Kitchawan Road, Room 34-225, Yorktown Heights, NY, 10598, United States, pharsha@us.ibm.com 1 - Hot Sales Logistics Optimization For ET Ba ak Erman, Bilkent University, Ankara, Turkey, basakerman@gmail.com Zeynep Yaprak Be ik, Deniz Berfin Karakoç, Umut Müdüro Lu, Yekta Jehat Mizrakli, Egehan Yanik ET is one of the leading food manufacturer in Turkey. Additonal to standard distribution system, ET has a local distribution system that enables trucks to visit smaller retailers and pursue hot sales. The aim of the project is to increase the efficiency of hot sales where demand by the retailers are better satisfied. The delivery route is divided into two: route from the main depot, which is far from customers, to the customers, and the route between customers. This project aims to maximize the utility of time spend in routes by assigning customers to trucks and identifying depot locations. 2 - Regularized Linear Regression via Robust Optimization Lens Hari Bandi, Massachusetts Institute of Technology, Cambridge, MA, United States, hbandi@mit.edu, Garud Iyender, Vineet Goyal There has been research in recent years to understand why regularized linear regression methods work well in the presence of noise. This problem has been approached by establishing relationship between robust optimization and regularized linear regression methods. In this work, we seek to understand the same for general loss functions used widely in Statistics, Machine Learning and Econometrics literature and we propose principled approaches to select regularization functions in order to optimally balance the bias-variance trade-off in regularized regression. 3 - On Comprehensive Mass Spectrometry Data Analysis For Quality Assessment Of Biological Samples Sameer Manchanda, Purdue University, West Lafayette, IL, United States, Mikaela Meyer, Nan Kong, Qianqian Li, Yan Li Mass spectrometers have become promising instruments to acquire proteomic information, creating a need for a data analysis platform for classification of mass spectra and identification of important biomarkers. To meet this need, we present a comprehensive pattern recognition platform for spectrum preprocessing and classification. In a case study, the platform achieves higher than 90% sensitivity and specificity in distinguishing rat blood samples stored for different amounts of time and derives fingerprint patterns of serum proteins that are strongly associated with the sample classification. SC08 103A-MCC Undergraduate OR Prize - II Invited: Undergraduate Operations Research Prize Invited Session

71