Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TD62

and overstate outflows; we also show the probability distribution for the number of undocumented immigrants based on simulating our model over parameter value ranges. Our conservative estimate is 16.7 million for 2016, nearly fifty percent higher than the most prominent current estimate of 11.3 million, which is based on survey data and thus different sources and methods. The mean estimate based on our simulation analysis is 22.1 million, essentially double the current widely accepted estimate. n TD62 West Bldg 103A Joint Session DM/Practice Curated: Data Science and Analytics for Business Decision-Making Sponsored: Data Mining Sponsored Session Chair: Durai Sundaramoorthi, Washington University in Saint Louis, Saint Louis, MO, 63131, United States 1 - Multistakeholder Recommendation with Provider Constraints Edward Carl Malthouse, PhD, Northwestern, 1845 Sheridan Road, Evanston, IL, 60208, United States, Ozge Surer, Robin Burke Recommender systems optimize the utility of the user, but often there are other stakeholders. One setting is multisided platforms, which match buyers and sellers and it is necessary to jointly optimize the value for both. We propose an integer programming model, where constraints reflect the goals of the different stakeholders. Our model is a post-processing step, so it can easily be added to an existing recsys to make it multistakeholder aware. For computational tractability, we reformulate the IP using the Lagrangian dual and subgradient optimization. We use two data sets to evaluate the utilities of buyers and sellers and show that our approximation can achieve good upper and lower bounds. 2 - Frequently Bought Together: Market Basket Analysis with Shortened Web Link Click Data Christopher M. Smith, Air Force Institute of Technology, 2795 Ridge View Ct, Xenia, OH, 45385, United States, James Gallagher Social Media use has grown tremendously since the advent of “Web 2.0ö. To facilitate the sharing of information, link shortening has become a common method for fitting long web links into space-constrained social media posts. Additionally, most link shortening services provide analytic feedback to users such as how many clicks the link generated or a breakdown by location of those clicks. Market basket analysis using graph mining techniques was applied to this web link data to provide insight into an area’s information dissemination patterns. Patterns of access, including by hardware platform and referral sources, within clusters of web domains provide feedback to market research analysts. 3 - Pick-up, Delivery, or Both? An Online Grocer’s Optimal Fulfillment Models Chloe Kim Glaeser, Assistant Professor, UNC Kenan-Flagler Business School, 300 Kenan Center Drive, Chapel Hill, NC 27599, Chapel Hill, NC, United States, Xuanming Su, Ken Moon We partner with an online grocery retailer to answer the practice-based question of the optimal mix of delivery zones and fulfillment models using data-driven analytics. We investigate (A) how consumers respond to the locally tailored fulfillment options made available to them by the online grocer, and (B) how to leverage data to customize locally available fulfillment options while scaling the retailer’s operations. We employ a regression discontinuity design to find the effect of delivery introduction. Based on this empirical evidence, we build and estimate a structural model and perform a counter-factual analysis to estimate the revenue increase from additionally offering delivery. 4 - Machine Learning Approaches to Modeling Category Sales: Implications for Optimal Store-Level Pricing and Promotions Durai Sundaramoorthi, Washington University in Saint Louis, 10352 Conway Road, Saint Louis, MO, 63131, United States We propose machine learning approaches - Regression Trees (RT), Bagging (Bag), Random Forests (RF), and Boosted Trees (BT) - with modified loss functions as candidate predictive models of product category profit as functions of marketing variables (price, display, feature, price promotion), both within and across categories at the same retail store. Using store-level weekly scanner data from 24 product categories in each of 9 stores of a supermarket chain over a period of 5 years, we estimate these machine learning models and compare their predictive performance on a validation set. We find that these models with modified loss functions are meaningful in the optimization context.

n TD60 West Bldg 102B HAS Session #II Sponsored: Health Applications Sponsored Session Chair: Bryan A. Norman, Texas Tech University, Lubbock, TX, 79409-3061, United States 1 - Multi-appointment Patient Scheduling for Chemotherapy Scheduling Maryam Keshtzari, Texas Tech University, Texas, Lubbock, TX, United States, Bryan A. Norman Cancer clinics are often overwhelmed with a large number of patients requesting chemotherapy treatments. Many patients prefer to have their chemotherapy session on the same day they visit the oncologist. In this study a simulation-based optimization model is presented to create a more balanced schedule to improve patient wait time and reduce nurse workload stress. Different scenarios based on patient’s acuity level and length of the infusion appointment are designed and evaluated considering uncertainties related to cancellations and no-shows. 2 - Moving on Up: Appointment Rescheduling to Improve Outpatient Clinic Appointment Slot Utilization Shannon Harris, The Ohio State University, 1262 Eastwood Ave, Columbus, OH, 43203, United States, Bjorn Berg, Jerrold H. May, Luis G. Vargas, Nathan C. Craig It is a common assumption that outpatient clinics are passive participants in the appointment booking process. If a patient cancels an appointment during the lead time, and another patient does not book an appointment in the cancelled time slot, the slot will remain empty. However, if the clinic can actively manage the appointment bookings, it may be able to increase the expected utilization of its appointment slots by potentially rescheduling appointments during a scheduling horizon. We prescribe general rules for when and how such appointment movements should occur in order to maximize expected utilization. 3 - Managing Interruptions in Appointment Schedules in Physician Clinics Ali Dogru, University of Southern Mississippi, Hattiesburg, MS, 35406, United States, Sharif Melouk Physician clinics often encounter interruptions that affect their operations and impact the patient experience. Thus, in this research, we develop an interruption management procedure employing real-time patient notification. We use stochastic optimization to determine optimal appointment intervals and simulation optimization to establish a notification policy. Experimentation provides managerial insights. 4 - Outpatient Clinic Appointment Scheduling using a Multi-objective Table-input Simulation-optimization Approach Mohammad Dehghanimohammadabadi, Northeastern University, 170 Brookline Avenue, Unit 1025, Boston, MA, 02115, United States, Mandana Rezaeiahari Appointment scheduling (AS) is one of the key factors to enhance the patient satisfaction in healthcare services. A practical and robust appointment scheduling pattern allows clinics to utilize medical assets, equipment, and resources in an efficient manner. In this study, a multi-objective simulation-optimization (MSO) approach is applied to determine the most preferred appointment scheduling pattern for an outpatient clinic system with stochastic parameters. The developed MSO model is using the concept of table-experiment (appointments table) in a simulation environment (Simio) which is improved with an iterative metaheuristic algorithm (in MATLAB). n TD61 West Bldg 102C The Number of Undocumented Immigrants in the United States Emerging Topic: INFORMS Special Sessions Emerging Topic Session Chair: Mohammad Fazel-Zarandi, Massachusetts Institute of Technology, MIT, Cambridge, MA, United States 1 - The Number of Undocumented Immigrants in the United States Mohammad Fazel-Zarandi, Massachusetts Institute of Technology, MIT, Cambridge, MA, United States, Jonathan S. Feinstein, Edward H. Kaplan We apply standard operational principles of inflows and outflows to estimate the number of undocumented immigrants in the United States, using the best available data, including some that have only recently become available. We develop an estimate of the number of undocumented immigrants based on parameter values that tend to underestimate undocumented immigrant inflows

369