INFORMS 2021 Program Book

INFORMS Anaheim 2021

MD16

1 - Digital Platforms and Echo Chambers: A Comparison of News Platforms and Social Media Derya Ipek Eroglu, Virginia Tech, Blacksburg, VA, 24060-8034, United States, Onur Seref, Michelle Seref Social media enables people to learn about and express their reactions to the current news. Additionally, news platforms provide their readers with a digital public space for expressing their opinions. These platforms differ from each other in their news coverage. When combined with the public digital space provided, these platforms enhance echo chambers in which readers with similar ideological perspectives interact. We expect a higher echo chamber effect on news platforms than social media platforms. We select a sample of the news platforms included in AllSides.com, which covers media bias on a wide spectrum of outlets. We collect news and comment data from these platforms to analyze their biases. We collect data from Twitter, Reddit, and Gab to compare these platforms to social media. We use an automated discourse analysis framework in our study and present our findings. 2 - Analyzing the Dynamic Nature of Firm to Consumer Engagement on Social Media Sae Hoon Chang, PhD Student, Queen’s University, Kingston, ON, Canada, Ceren Kolsarici We investigate dynamics of firm-consumer interactions on social media platforms. Using sentiment analysis and multiple regularization processes, we examine Tweet characteristics in firm-to-many and firm-to-one communications that lead to different levels of consumer engagement. The results suggest that emotional content in mass engagement, and precision and clarity of the message in individual engagement stages lead to higher likes and retweets. MD16 CC Room 201D In Person: Healthcare Analytics, During the Pandemic and Beyond General Session Chair: Sara Nourazari, California State University, Long Beach, CA, 92648-0906, United States 1 - Using Surgical Schedule Bed Board Modeling Results from Pandemic for Planning Future Hospital Occupancy Franklin Dexter, Professor, University of Iowa, Department of Anesthesia, Division of Management C, Iowa City, IA, 52242, United States, Richard Epstein, Pengyi Shi When the hospital census is high, perioperative medical directors or operating room managers may need to postpone some surgical cases scheduled within a few (e.g., <3) workdays. For the COVID-19 pandemic, we used data from state database and detailed data from a large hospital. Monte-Carlo simulations and time series analyses showed that, for purposes of comparing procedures at the same hospital, there is no loss of information by summarizing the probability distributions of hospital length of stay for elective surgical cases using single numbers, the percentages of cases among patients staying longer than overnight. This finding simplifies the mathematics for constructing dashboards or summaries of information system data to help the medical director make decisions. 2 - Machine Learning and Clustering-based Approach for County- level Covid-19 Analysis Charles D. Nicholson, University of Oklahoma, Norman, OK, 73019, United States COVID-19 is a global pandemic threatening the lives and livelihood of millions of people across the world. Due to its novelty and quick spread, scientists have had difficulty in creating accurate forecasts for this disease. This is especially true for regionally specific predictive models due to either limited case histories or other unique factors characterizing the region. This paper employs both supervised and unsupervised methods to identify the critical county-level demographic, mobility, weather, medical capacity, and health-related county-level factors for studying COVID-19 propagation. We use this feature subspace to aggregate counties into meaningful clusters to support more refined disease analysis efforts. 3 - Early Detection of Trend Shifts in Emergency Department Surges During the Covid-19 Pandemic Sara Nourazari, California State University, Long Beach, Long Beach, CA, 92648-0906, United States, Samuel Davis, Rachel Granovsky, Dean J. Straff, Joshua W. Joseph, Leon D. Sanchez A change detection tool is developed for tracking and early detection of trend shifts to help identify patterns in volume surges and declines in emergency departments during and after the COVID-19 pandemic in the U.S. This will allow studying the impact of different state-level and national guidelines and strategies on COVID-19 related restrictions and their downstream effects. At a macro level, this method can help study the impact of the pandemic on population health and emerging patterns of specific conditions such as mental health.

minimizing the costs of inventory, backorders and production. We use a Deep Reinforcement Learning (DRL) methodology to find solutions. Larger problem instances encounter some challenges, which we resolve by utilizing domain knowledge to support the DRL algorithm.

MD14 CC Room 201B In Person: Advances in Machine Learning and Optimization Methods General Session Chair: Petros Xanthopoulos, Stetson University, DeLand, FL, 32723, United States 1 - Scenario-based Robust Optimization for Decision-making under Binary Uncertainty Kai Wang, Carnegie Mellon University, Pittsburgh, PA, 02215- 4212, United States, Alexandre Jacquillat, Mehmet Aydemir This paper addresses data-driven optimization problems under categorical uncertainty. We propose a scenario-based robust optimization approach that combines stochastic programming (by constructing scenarios), robust optimization (by building discrete uncertainty sets), and data-driven optimization (by defining scenarios and uncertainty sets from machine learning classification outputs). We implement it on vehicle routing problem and ambulance dispatching problem. Results suggest that our approach outperforms benchmarks based on deterministic, stochastic, and robust optimization. 2 - Unsupervised Ensemble Learning Based on Internal Quality Measures and Modern Portfolio Theory Petros Xanthopoulos, Stetson University, DeLand, FL, 32723, United States Unsupervised ensemble learning or consensus clustering is the process of combining multiple clustering solutions into one with more robust characteristics. In this presentation we propose an weighted consensus clustering approach based on internal quality measures. We demonstrate the its advantages in terms of clustering quality and robustness. We also describe its performance improvement in terms of number of cluster identification. We present a tweak based on modern portfolio theory that allows to control the algorithmic robustness through a simple optimization model. 3 - Non-overlapping Group Structured Sparsity Problems: Computations Miju Ahn, Southern Methodist University, Dallas, TX, 75205, United States We introduce a new formulation and an iterative algorithm for a group-sparse representation problem where subsets of model variables form non-overlapping groups. The proposed algorithm solves a reweighted group lasso at each iteration, and computes a directional stationary solution which achieves global optimality under some conditions. We present results of numerical experiments conducted with synthetic and real datasets showing that our method achieves superior performance compared existing methods in many criteria including prediction accuracy, relative error, and group recovery success rate. 4 - Logical Sequential Pattern Mining and Classification via Mixed-integer Optimization Ruilin Ouyang, Northeastern University, Boston, MA, 02120, United States, Chun-An Chou Mining temporal patterns in times series data is an important data mining task in various application areas. However, the conversion of time series data into symbolic data is often required a priori. Moreover, it remains a challenging task to mine critical temporal patterns which are discriminative with explicit time information. In this work, we present a novel mixed integer linear programming model to optimize a set of logical sequential patterns with a maximum coverage of samples in target class by considering critical patterns appear synchronously among all time-series samples. Furthermore, we propose an efficient algorithm to solve the proposed model in a short run time. Finally, we demonstrate the effectiveness of the proposed methods on both simulated and real datasets comparing with the state-of-art sequential pattern mining method. MD15 CC Room 201C In Person: Data Analytics in Social Media and Information Systems General Session Chair: Derya Ipek Eroglu, Virginia Tech, Blacksburg, VA, 24060-8034, United States

77