2019 Student Poster Presentations

*presenting author listed

 

Student Name Email Affiliation Advisor Title Abstract
Kehinde Odubela kaodubela@aggies.ncat.edu North Carolina A&T State University Dr. Steven Jian, Dr. Lauren Davis, Ritson Delpish Effect of Underlying Factors on Food Distribution Forecasts Using  Visual Analytics In this research, we used Visual Analytics (VA) to study the effect of certain underlying factors on the forecast generated for food distribution by a local hunger relief organization. Specifically, we used already tested forecasting techniques to predict the expected quantity of distributed food for the underlying factors identified.
Behshad Lahijanian b.lahijanian@ufl.edu University of Florida Dr. Michelle Alvarado Chance-constrained Stochastic Programming model for Reducing Hospital Readmissions In response to the Hospital Readmission Reduction Program (HRRP), hospitals are seeking care strategies to reduce their readmission rates. We develop a multi-condition care strategy model with probabilistic constraints to help hospitals prioritize treatment plans. The hospital aims to identify care strategies that  provide a confidence level that the target readmission probability will be achieved. We explored the trade- off between the cost of care, probability of readmission, and confidence levels.
Yeawon Yoo yyoo12@asu.edu Arizona State University Dr. Adolfo Escobedo A new binary programming formulation and social choice property for Kemeny ranking aggregation We introduce a binary programming formulation for the generalized Kemeny ranking aggregation problem. The new formulation, which is connected with the weak order polytope, provides comparative advantages including reduced memory requirements and faster computing times. Moreover, we develop a new social choice property called the Generalized Condorcet Criterion. The property is leveraged to develop a structural decomposition algorithm, through which certain large instances of the NP-hard Kemeny aggregation problem could be solved exactly in a practical amount of time.
Brittany Green greenb5@mail.uc.edu University of Cincinnati Investigating Donation Uncertainty in Not-for-Profit Organizations Not-for-profit organizations can face considerable uncertainty of donations and funding. To alleviate this challenge, some not-for-profit organizations also use for-profit activities to support and subsidize the mission of the organization. This work investigates trade-offs regarding the critical decision of determining the amount to allocate to selling versus donating under uncertainty. We approach the solution to this problem by using robust optimization and Monte-Carlo simulation to identify optimal policies for not-for-profit organizations.
Shenghan Guo sg888@scarletmail.rutgers.edu Rutgers Spatial-Temporal Modeling and Monitoring of Melt Pool Evolution in Laser-Based Additive Manufacturing This work develops a spatial-temporal model for the melt-pool thermal images collected during the additive manufacturing of Ti-6Al-4V thin walls. Spatial-Temporal Conditional Autoregressive (STCAR) model underlies the framework. A hierarchical monitoring system is innovated on top of the model, in which a sequence of Level I control charts and a Level II control chart simultaneously capture the local and global variations.
Francisco Castillo Zunino fj.castillo.zunino@gatech.edu Georgia Tech Pinar Keskinocak Approximation Algorithms for the Multiple Knapsack Problem with  Grouped Items Motivated by patient scheduling, the multiple knapsack problem with grouped items maximizes profits by assigning items to knapsacks, without exceeding capacities. Items are partitioned into groups where either all items in a group are assigned or none. We propose three algorithms that each solves a: linear relaxation, knapsack problem, and multi-dimensional knapsack problem. Solutions’ profits are not less than the optimal, but can exceed capacities. We test the performance on random instances.
Isabella Thavi Sanders isabella@gatech.edu  Georgia Tech Strategic Market Deployment Planning: Farm-to-Table Platforms Development of a data-driven market deployment planning methodology utilizing a mixed integer program, towards applicability in the context of farm-to-table logistics platforms, to create a roadmap consisting of sets of markets targeted for deployment at each time-phase of business development. This roadmap is optimized for expected profit, allowing companies to achieve market expansion plans and profit objectives under limited resources and managed risks.
Brittany Segundo brittany.segundo@tamu.edu Texas A&M Lewis Ntaimo Real-time Decision Making During Large-Scale Wildfires with Stochastic Programming Large-scale wildfires escaping initial containment efforts require a coordinated response from numerous agencies. While stochastic programming has been used to pre-position firefighting resources, we employ a multistage stochastic program to inform the allocation of resources during the course of a wildfire. Decision-makers face volatile fire growth as a result of uncertain weather changes, but previous response efforts also impact fire progression. We approximate this multistage program with endogenous uncertainty using rolling horizon optimization.
Jiangye Gong gongjiangyue@gmail.com Texas A&M Lewis Ntaimo Pathways Modeling and Scheduling for Connected Community Health under  Uncertainty Pathways are structured and time-framed guidelines used by health-related services to detail essential steps of a process with a specific outcome. In this talk, we present a stochastic programming model for connected community health to schedule clients and resources for each step of the pathway under uncertainty in resource availability. The goal of the model is to optimize clients wait-time while providing workload balance among resources of the same type.
Temitayo Ajayi ta21@rice.edu Rice University Andrew Schaefer Objective Selection for Prostate Cancer Treatment: A Generalized Inverse Optimization Approach Inverse optimization is an emerging method in data science and has been applied to radiation therapy planning to weight multiple conflicting objectives. However, it is also important to deliver a parsimonious model, which leads to objective selection. We provide the first application of inverse optimization for objective selection. We apply our methods to prostate cancer treatment, and we show that this is a reliable way to recover clinical preferences and help standardize the planning process.
Shubhra Paul spaul1@aggies.ncat.edu North Carolina A&T State University Dr. Lauren B. Davis Predicting Food Donor Contribution Behavior Using Support Vector Regression Hunger and food insecurity are present in each American county. Food banks are nonprofit hunger relief organizations that collect donations from donors and distribute food to local agencies which serve people in need. The uncertainty of donation is a challenge for food banks. In this research, we analyze local food bank donation data and propose a predictive model to forecast the contribution of different donors with Support Vector Regression. Our study shows the necessary behavioral attributes to classify donors and the best way to cluster donor data to improve the prediction model.
Vinayak S. Ahluwalia vahluw@umich.edu University of Michigan Brian Denton A Branch-and-Bound Approach for Infinite-Horizon MDPs and its Application to Medical Decision-Making Markov decision processes (MDPs) subject to parameter ambiguity may lead to conflicting recommendations, especially in disease treatment. A study of this problem for finite-horizon MDPs succeeded with a branch-and-bound approach, so we extend the study to infinite-horizon MDPs and recommend algorithmic designs that minimize computation time. We demonstrate our algorithm's effectiveness over standard methods through random test instances and a previously published model for HIV treatment.
Sofia Perez-Guzman perezs@rpi.edu RPI A Two-Stage Stochastic Model for the Allocation of Critical Goods in a Post-Disaster Scenario The dynamic nature and uncertainty of disaster relief scenarios have been modeled with respect to aid availability. The allocation of aid also depends on the population in need and their location, which is not deterministic since it changes while relief occurs. This work develops a two-stage stochastic optimization model that incorporates human suffering through deprivation cost functions, and variation of demand of aid through a linearization a Lotka-Volterra model approximation. A study case is analyzed.
Rosemarie Santa-González rosemarie.santa@gerad.ca Universite du Quebec a Montreal The Multi-Period Location Routing Problem:  A Mobile Clinic Application This study illustrates the Multi-Period Location Routing Problem with a Mobile Clinic (MC) application. MCs allow healthcare practitioners to provide services to populations with limited access to healthcare. In MC operations depots, locations, and schedules must be selected. The proposed model is tested on real-life data from Premier Urgence Internationale.
Claudia Ramirez cramir16@uncc.edu University of North Carolina Charlotte Mathematical Optimization for Habitable Layout Design driven by Psychophysical Criteria This research aims at designing a spacecraft layout for a mission to Mars to minimize volume and maximize crewmembers’ health, performance, and vehicle safety. As part of the design process we evaluated designs by correlating psychophysical factors to proximity and functionality of task volumes. Relationship between physical arrangements and psychophysical satisfaction of needs provides greater insight in the design process to achieve optimal results.
Sefakor Fianu sfianu@aggies.ncat.edu North Carolina A & T State University  Nested Markov Decision Process Model for Equitable and Effective Supply Distribution Food banks have evolved from backyard operations into complex network organizations. Distribution decisions made by food banks include transfer of supplies from branch to branch and distribution to partner agencies. This problem is formulated as a Nested-Markov Decision Process to find optimal distribution policies that minimizes unsatisfied demand and waste.
Joseph Yeiter jny3488@g.rit.edu RIT ruben Proano Vaccine Procurement Over Time Global immunization remains at 85% with an estimated 1.5 million deaths each year due to preventable diseases. Increasing immunization rates remains unattainable due to vaccine affordability. Strategies have been utilized to improve affordability for some countries, though overall market affordability still suffers. This study examines procurement strategies in a coordinated global vaccine market where buyers have incentive to buy early. We seek to understand whether the global vaccine procurement cycle could be scheduled in a way that benefits both manufacturers and buyers.
Bruno alves Maciel  ba8641@rit.edu RIT ruben Proano  Optimized pricing strategies for vaccine affordability and profits A 3-stage optimization process shows how purchasing vaccines with a lump-sum price for a bundle of products in a partially coordinated market affects affordability and profits when compared to a price per dose of vaccine. Results suggest that lump-sum pricing redistributes savings among countries that coordinate their negotiation, increasing affordability without driving profits below the expected return on investment. 
Shubhra Paul spaul1@aggies.ncat.edu NCAT Lauren Davis Predicting Food Donor Contribution Behavior Using Support Vector Regression Hunger and food insecurity are present in each American county. Food banks are nonprofit hunger relief organizations that collect donations from donors and distribute food to local agencies which serve people in need. The uncertainty of donation is a challenge for food banks. In this research, we analyze local food bank donation data and propose a predictive model to forecast the contribution of different donors with Support Vector Regression. Our study shows the necessary behavioral attributes to classify donors and the best way to cluster donor data to improve the prediction model.
Veronica White vmwhite@wisc.edu University of Wisconsin Madison Gabriel Zallas-Caban Utilizing causal inference and hypothesis testing to evaluate and advise a local policing initiative  The Madison Addiction Recovery Initiative (MARI) is a Smart Policing Initiative in the City of Madison and Dane County, Wisconsin, that seeks to offer persons who are stopped for low-level, victimless offenses an alternative to jail via medication-assisted and behavioral drug treatment.  Using individual-level data, collected over a three year period, causal inference and hypothesis testing are utilized to discover key insights into future policy decisions for the MARI program.  
Jusse Aline Hirwa Jhirwa@mymail.mines.edu Colorado School of Mines  Alexandra Newman Combined heat and power as an alternative energy supply for South Africa’s industry sector Lack of access to reliable energy is a major concern for firms in South Africa’s industry sector. As a result, they suffer losses that cripple the region’s economic growth and development. Most firms must consider alternative means of energy supply.  Combined heat and power (CHP) can be one solution. We employ an optimization model to determine the conditions for which CHP is cost-effective while meeting total energy needs for the mining, and steel and iron sectors. 
Lois Kamga-Ngameni lkamgang@mymail.mines.edu Colorado School of Mines  Alexandra Newman Historically, Darcy’s equation has been used to predict hydrocarbon production, but is inaccurate in unconventional reservoirs.  We (i) present the more accurate Barree-Conway model (2004); (ii) use regression to develop better predictions of fracture conductivity (the rate at which fluid flow occurs in unconventional reservoirs) as a function of fracture pressure and permeability of reservoir rock; and (iii) suggest corresponding optimization models that improve production efficiency
Oluwaseun Ogunmodede oogunmod@mines.edu Colorado School of Mines  Alexandra Newman Using Renewables towards Net-Zero System Planning Renewable energy technologies are becoming increasingly important due to diminishing supplies of conventional non-renewable resources. We demonstrate the capabilities of an optimization model that minimizes costs and carbon emissions, while adhering to system sizing demands for a variety of venues, e.g., suburban homes or military installations. Given the size of realistic instances, we propose solution-expediting techniques to our mixed-integer program to solve under the time limits required.
Justice Darko jdarko@aggies.ncat.edu North Carolina A & T State University Hioshin Park A Dynamic Transit Model for Vulnerable road users Vulnerable road users are most sensitive to risk and uncertainties in transit networks. These travelers usually prefer decisions that is one-shot, avoiding high-risk actions. We model the behavior of the constant risk-sensitivity of these travelers by assuming the exponential function. Integrating the exponential function into the Markov Decision Process ensured that the actions taken by the traveler is now conditioned on the level of uncertainty and risk in the stage of decision making.