2015 Informs Annual Meeting

SA31

INFORMS Philadelphia – 2015

SA30 30-Room 407, Marriott Research from 2015 Richard E. Rosenthal Early Career Connection Program Participants Sponsor: Analytics Sponsored Session Chair: Aurelie Thiele, Lehigh University, 200 W Packer Ave, Bethlehem, PA, 18015, United States of America, aut204@lehigh.edu 1 - Overview of “The Richard E. Rosenthal Early Career Connection Program” Aurelie Thiele, Lehigh University, 200 W Packer Ave, Bethlehem, PA, 18015, United States of America, aut204@lehigh.edu This short talk will provide an overview of the Richard E. Rosenthal Early Career Connecting Program, held in conjunction with the yearly Analytics conference in the spring. It will focus on the 2015 edition of the program, co-organized by Michelle Opp of SAS and myself. 2 - A Protein Scoring Function using Support Vector Machine Shokoufeh Mirzaei, Cal Poly Pomona, 3801 West Temple Avenue, Pomona, CA, 91768, United States of America, smirzaei@cpp.edu, Silvia Crivelli In this paper a knowledge-based scoring function for quality assessment of protein decoy models is developed. To this end, a benchmark data set from CASP 8, 9 and 10 is used. The dataset includes measurements of proteins structural features that are seemingly having significant impacts on the quality of predicted structures. 3 - Biologically-guided Radiotherapy Treatment Plan Optimization Ehsan Salari, Wichita State University, 1845 Fairmount St, Wichita, KS, Ehsan.Salari@wichita.edu Radiotherapy treatments are delivered in daily fractions over the course of one to several weeks. There is clinical evidence suggesting that patients with specific tumor sites may benefit from delivering larger radiation doses in fewer fractions. However, current treatment regimens use a fixed radiotherapy plan in all fractions. This research aims at developing a spatiotemporal planning approach that allows to investigate the potential benefit of temporal variation in the plan across fractions. SA31 31-Room 408, Marriott Mathematical Optimization Models for Data Science Chair: Dolores Romero Morales, Copenhagen Business School, Porcelaenshaven 16 A, Frederiksberg, DK-2000, Denmark, drm.eco@cbs.dk 1 - Learning Tailored Risk Scores from Large Scale Datasets Berk Ustun, PhD Candidate, MIT, 20 Highland Avenue Apt. 2, Cambridge, MA, 02139, United States of America, ustunb@mit.edu, Cynthia Rudin Risk scores are simple models that let user assess risk by adding, subtracting and multiplying a few small numbers. These models are widely used in medicine and crime prediction but difficult to learn from data because they need to be accurate, sparse, and use integer coefficients. We formulate the risk score problem as a MINLP, and present a cutting-plane algorithm to solve it for datasets with large sample sizes. We use our approach to create tailored risk scores for recidivism prediction. 2 - A Multi-objective Approach to Visualize Proportions and a Binary Relation by Rectangular Maps Dolores Romero Morales, Copenhagen Business School, Porcelaenshaven 16 A, Frederiksberg, DK-2000, Denmark, drm.eco@cbs.dk, Emilio Carrizosa, Vanesa Guerrero We address the problem of representing individuals, to which there are proportions attached and a binary relationship, by means of a rectangular map, i.e., a subdivision of a rectangle into rectangular portions, so that each portion is associated with one individual, the areas of the portions reflect the proportions, and portions adjacencies reflect adjacencies in the binary relationship. We formulate this as a three-objective Mixed Integer Nonlinear Problem and numerical results are presented. Sponsor: Data Mining Sponsored Session

2 - Inapproximability of Truthful Mechanisms via Generalizations of the VC Dimension Gal Shahaf, PhD Candidate, Hebrew University, Nataf 63, Nataf, 9080400, Israel, gal.shahaf@mail.huji.ac.il, Amit Daniely, Michael Schapira Algorithmic mechanism design (AMD) studies the delicate interplay between computational efficiency, truthfulness, and optimality. We focus on AMD’s paradigmatic problem: combinatorial auctions, and present new inapproximability results for truthful mechanisms in this scenario. Our main technique is a generalization of the classical VC dimension and the corresponding Sauer-Shelah Lemma. Joint work with Amit Daniely and Michael Schapira 3 - Efficient Procurement Auctions with Increasing Returns Lawrence Ausubel, Professor of Economics, University of Maryland, Department of Economics, College Park, MD, 20742- 7211, United States of America, ausubel@econ.umd.edu, Christina Aperjis, Oleg Baranov, Thayer Morrill For procuring from sellers with decreasing returns (or selling to buyers with diminishing marginal values), there are known efficient dynamic auction formats. In this paper, we report progress in designing an efficient dynamic procurement auction for the case where bidders have increasing returns. The auctioneer names a price, and bidders report the minimum and maximum quantities that they would sell at that price. The process repeats with lower prices, until the efficient outcome is discovered. 4 - Competing Combinatorial Auctions

Marion Ott, RWTH Aachen University, Templergraben 64, Aachen, 52062, Germany, marion.ott@rwth-aachen.de, Thomas Kittsteiner, Richard Steinberg

What is the benefit of an auction format that allows for package bids for a seller who wants to sell a set of distinct items? We show that the answer depends on whether a seller faces competition from another seller. For a simple, tractable model we give conditions under which a seller with the choice between VCG mechanisms with or without package bidding prefers to disallow package bidding if another seller with the same options is present.

SA29 29-Room 406, Marriott Analytics

Sponsor: Analytics Sponsored Session Chair: Harrison Schramm, Navy Headquarters Staff, 1507 22nd Street South, Arlington, VA, 22202, United States of America, Harrison.Schramm@gmail.com 1 - Identifying Shortfalls in Library Holdings through Analysis of References in Faculty Publications Ziyi Kang, University of Pittsburgh, 1048 Benedum Hall, Department of Industrial Engineering, Pittsburgh, PA, 15261, United States of America, zik3@pitt.edu, Shi Tang, Louis Luangkesorn, Fan Zhang, Yunjie Zhang, Berenika Webster University libraries measure their contribution to research in part through providing reference material cited by faculty in their publications. One difficulty is that article references are often abbreviated in non-standard ways. To compare references with library holdings we apply text processing methods such as normalization, string distances, and word splitting to determine if a reference is held by the library. We apply this to one subject area and validate the accuracy of the method. 2 - Assessing the Effects of Cross-Season Fairness Scheme on the Competitive Balance of NFL Schedules Niraj Pandey, University at Buffalo, 342 Bell Hall, North Campus, Buffalo, NY, 14260, United States of America, npandey@buffalo.edu, Murat Kurt, Mark Karwan, Kyle Cunningham The National Football League (NFL) is the highest revenue generating sports league in the world. Although the league’s scheduling routine has evolved over the years to ensure fairness, recent schedules exhibit significant imbalances in several dimensions, particularly in teams’ rest durations between games. We develop a two-phase MILP approach to create fairer schedules and evaluate the price of the league’s practice of rotating venues of the games on a multi-year basis on their competitiveness.

47