Informs Annual Meeting 2017

TD61

INFORMS Houston – 2017

2 - Teaching and Learning of Economic Decisions for Uncertain and Complex Projects K. Jo Min, Iowa State University, 3004 Black Engineering, Department of Industrial &, Ames, IA, 50011-2164, United States, jomin@iastate.edu, John Jackman, Michelle Zugg Economic decisions for projects that are complex and facing uncertainties are inherently difficult. In this paper, based on simple principles of stochastic optimal control, we show such subject can be taught and learned in an undergraduate course. Empirical statistical analysis of student learning outcomes as well as insights from experience will be presented. 3 - Fun and Innovative Ways to Teach Computer Programming Concepts Stephany Coffman-Wolph, Lecturer, The University of Texas- Austin, 2880 Donnell Drive, Round Rock, TX, 78664, United States, scoffmanwolph@utexas.edu Although most college students are familiar with computers, almost all stem majors must take one or more introductory computer courses. Teaching these classes can be challenging as students are unaware of what being a computer scientist really entails. The activities presented are designed to be extremely cost- effective, easily portable, require minimum preparation, and do not require a lab or prior computer experience. The end goal of these activities is to introduce various computer science topics using familiar everyday experiences and fun physical activities. 4 - Reimagining the Engineering Textbook: Learning-centered Textbook Design Kingsley Anthony Reeves, Associate Professor, University of South Florida, 4202 E. Fowler Avenue, ENB118, Tampa, FL, 33620, United States, reeves@usf.edu, Grisselle Centeno, Garrett Bowleg, Brittany Clift We present the initial results of a design effort to reimagine the engineering textbook. The goal was to develop a Probability and Statistics for Engineers textbook that was designed with the needs of instructors, the needs of students, and learning theory in mind. The resultant design features will be displayed along with their explicit connection to best practices as identified in the teaching and learning literature. 370B Joint Session SpORts/Practice: Sports Analytics IV Sponsored: SpORts Sponsored Session Chair: Walter DeGrange, CANA Advisors, Chapel Hill, NC, 27517, United States, wdegrange@canallc.com 1 - Rank and Score: Athlete-event Assignment Optimization Matthew Bailey, Bucknell University, Bucknell University, Taylor Hall, Lewisburg, PA, 17837, United States, matt.bailey@bucknell.edu, Maciek A.Nowak With over 16,000 programs and one million secondary school athletes annually, track and field is second only to football in national participation. For coaches managing these large teams the focus is often solely on the training and development of athletes. As a result, coaches may overlook and underappreciate the competitive advantage of strategically assigning athletes to events to maximize their team’s points in a meet. Building upon previous work, we present a generalization and extension of an assignment problem where we simultaneously assign, rank, and score athletes in events to maximize the expected total team points scored under a variety of scenarios, league rules, and preferences. 2 - Sports Analytics: Viable Career Path or Something Really Smart People Made up to Try to Get in the Game With the recent success of sports teams heavily using analytics (Cubs, Patriots, Penguins, Warriors, Leicester City F.C.), does this mean that analytics has gained a foothold in the sports world? And if so, is there a career path that a high school student can use to become a sports analytics professional? This presentation attempts to answer both of these questions and summarizes all the areas of the application of analytics in sports. Walter DeGrange, Cana Advisors, Chapel Hill, NC, 27517, United States, wdegrange@canallc.com, Jesse McNulty TD61

3 - Predicting Collegiate Lacrosse Play Performance at the U.S. Major League Lacrosse Level Walter DeGrange, CANA Advisors, 6727 Falconbridge Rd, Chapel Hill, NC, 27517, United States, wdegrange@canallc.com, Jesse McNulty, Harrison Schramm Predicting the performance of amateur athletes at the professional level is challenging in any sport. Very few of the players get the opportunity to play professionally. In this presentation we will show the results of using machine learning and traditional techniques on a data set of over 45,000 collegiate lacrosse players that covers a three year period. The predictors are Major League Lacrosse shooting and defending data sets. Being able to connect data such as school, team strength of schedule, shooting, goal tending, and historical performance of professional players allows insight into the potential for a collegiate player’s performance at an elite level. 4 - Using Machine Learning to Improve Football Play Calling Effectiveness Jonathan David Lonski, Student, Clemson University, Clemson, SC, 29634, United States, jdlonsk@g.clemson.edu Using past data from NCAA football games, we developed a tool that assists coaches with making in-game decisions regarding plays to call. Historical data describing the down, distance to gain a first down, field position, offensive formation, defensive front, and other key indicators. Our tool’s objective is to recommend the play that has the highest expected average gain and the highest probability of achieving a first down, based on the situation and defensive formation being faced. 370C Developments in Bayesian Data Analysis Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Arman Sabbaghi, Purdue University, West Lafayette, IN, 47907, United States, sabbaghi@purdue.edu 1 - A Missing Data Perspective for Modeling Deformations in 3D Printed Products Sobambo Sosina, PhD, Deloitte University, MD, United States, bambososina@gmail.com In the past decade, additive manufacturing (AM), a layer-by-layer approach for printing 3D objects, has attracted a lot of attention from manufacturers. However, dimensional accuracy control currently poses a challenge in AM, often due to deformations resulting from phase changes in the printing material. Building on work done by Huang et al. (2015) on the minimization of deformations, we explore a missing data approach for understanding deformations, subsequently utilizing Bayesian models in the estimation of deformation, and compensation plan design for minimizing the deformation effect 2 - Designing for What’s Important: Combining Bayesian and General Weighted Optimality Criteria Jon Stallings, North Carolina State University, jwstalli@ncsu.edu How one compares experimental designs depends heavily on the analysis goals (minimizing estimation variance, estimation bias, prediction variance, etc.). If one decides to focus on minimizing estimation variance, there is still the question of what has to be estimated and whether some effects are more important to estimate. DuMouchel and Jones (1994) tackled this issue with factorial experiments through a Bayesian framework assigning relative importance through prior variances. The Bayesian optimal design minimizes a summary of posterior variances and hence focuses on efficient estimation of effects with large prior variance. Stallings and Morgan (2015) introduced a class of general weighted optimality criteria that also allows the experimenter to assign relative importance to model effects through weighted variances. In this talk I compare the relative advantages and disadvantages of the two approaches in terms of their interpretability and utility for tailoring experiments to a researcher’s goals. I then present a new criteria function that combines the two and demonstrate the resulting designs under multiple cases of different types of factorial experiments. 3 - A Bayesian Analysis of Communication Activity on Social Media Networks Vinayak Rao, PhD, Purdue University, West Lafayette, IN, 47907, United States, varao@purdue.edu Modeling communication activity among individuals on email, social media and citation networks is a topic of recent interest in statistics and machine learning. We examine such activity through the lens of two models: latent space models, where communication between individuals is determined by latent “interests”, and reciprocal models, where this is a function of previous communication. We find both such models are inadequate for a variety of real-world communication data sets, and propose new models that address these limitations. Approaches we describe incorporate message content, include hierarchies of senders and receivers, and well as latent spaces that modulate reciprocal behavior. TD62

398