Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TC62

3 - A Chance Constrained Model to Locate an Automated Central Fill for a Community Pharmacy Network Hamdy Salman, 1989, Pittsburgh, PA, 15216, United States, Wei Wang, Bo Zeng, Bryan A. Norman, Bryan A. Norman Community pharmacies play an important role in providing patient safety by providing Pharmacist To Patient (PTP) interaction. One of the strategies to provide more PTP interactions in a community pharmacy is to use an automated Central Fill. An automated central fill offloads part of the pharmacy’s prescription filling process. Our research focuses on locating a central fill that can provide relief for multiple pharmacies under stochastic demands. We provide a chance constrained capacitated facility location model when pharmacy demands are variable. The results include a comparison between using a chance constraint model and using a simpler policy to determine the central fill location. 4 - Sequential Modeling and Optimization for Resources in Mail-order Pharmacy Automation Systems Duaa Serhan, Binghamton Uinversity, 4400 VESTAL PKWY E, Binghamton, NY, United States, Sang Won Yoon, Husam Dauod This research proposes an integrated optimization strategy to improve resource planning and replenishment operations in Central Fill Pharmacies (CFPs). The objective of this research is to develop a resource planning solution that determines the number of required dispensers, canisters, and replenishment operators. An integer programming mathematical model is developed to solve the planogram design and replenishment planning problems simultaneously. Tabu Search and Genetic Algorithm are proposed to solve this NP-hard problem efficiently. This research is expected to help with understanding the factors that influence CFP resource planning. n TC61 West Bldg 102C Optimization Society Award II Sponsored: Optimization Sponsored Session Chair: David Morton, Northwestern University, IEMS Department, 2145 Sheridan Road, Evanston, IL, 60208, United States 1 - Learning Semidefinite Regularizers from Data Yong Sheng Soh, California Institute of Technology, Pasadena, CA, United States Regularization functions are widely used in optimization-based approaches for solving ill-posed inverse problems. These serve to induce desired structure in solutions, and are typically chosen based on domain-specific expertise. We consider the problem of learning regularizers from data in settings in which domain knowledge is not available. The regularizers we obtain are convex functions that can be computed via semidefinite programming. Our approach is based on computing certain structured factorizations of data matrices, and it requires the Operator Sinkhorn iteration as a subroutine. 2 - Pareto Efficiency in Robust Optimization Nikos Trichakis, MIT, Cambridge, MA, United States, Dan Andrei Iancu We formalize the concept of Pareto efficiency in the context of robust optimization (RO) for linear optimization. We argue that RO need not produce solutions that are Pareto robustly optimal (PRO), and illustrate how this could lead to suboptimal performance. We provide a theoretical characterization of PRO and show how to verify and generate PRO solutions by solving optimization problems of the same complexity as the underlying RO problems; hence, the potential improvements from PRO come at no extra computational cost. Numerical studies demonstrate PRO solutions to have significant upside. 3 - On the Construction of Converging Hierarchies for Polynomial Optimization Based on Certificates of Global Positivity Amir Ali Ahmadi, Princeton University, Dept. of Operations Research & Financial Eng., Sherrerd Hall (room 329), Charlton Street, Princeton, NJ, 08544, United States, Georgina Hall In recent years, techniques that combine semidefinite programming with Positivstellensatze from algebraic geometry to produce a converging hierarchy of lower bounds for polynomial optimization problems have gained much popularity. In this talk, we show that such hierarchies can in fact be designed from much more classical statements in algebra dating back to the 1920s. We also provide a converging hierarchy of lower bounds for polynomial optimization problems which does not require optimization at all, but simply the ability to multiply polynomials together.

4 - Optimization Society Award David Morton, Northwestern University, IEMS Department, 2145 Sheridan Road, Evanston, IL, 60208, United States Winners of the following Optimization Society Prizes will present their work: Student Paper Prize; Young Researchers Prize; and Farkas Prize. n TC62 West Bldg 103A Joint Session DM/Practice Curated: Optimization in Data Mining and Analytics Sponsored: Data Mining Sponsored Session Chair: Young Woong Park, Iowa State University, Ames, IA, 50011, United States 1 - A Mathematical Programming Approach for Imputation of Unknown Journal Ratings in a Combined Journal Quality List Young Woong Park, Iowa State University, 2167 Union Drive, 2139 Gerdin Business Building, Ames, IA, 50011, United States, Jinhak Kim, Alvin Williams Many journal quality lists rate limited numbers of journals. However, in many academic institutions, fair evaluation procedure is needed to rate journals that are not listed in their trusted journal quality lists. In this research, mathematical programming models are proposed to determine unknown ratings of multiple journal quality lists only using their known rating information. The objective of the models is to minimize the total number of instances where two journals are rated in opposite order by two quality lists. Computational results based on the journal quality list data in https://harzing.com show that our method outperforms existing multiple imputation algorithms. 2 - Recent Advances in Mixed-integer Optimization Approaches to Feature Subset Selection Yuichi Takano, University of Tsukuba, Tsukuba-shi, Japan, Ryuhei Miyashiro Mixed-integer optimization (MIO) approach to feature subset selection was proposed in the 1970s, and recently it has received renewed attention due to progress in algorithms and hardware. In this talk, I review some recent advances in MIO approaches to feature subset selection and introduce latest research developments of our group. 3 - Optimal Clustering on a Graph GOKCE Kahvecioglu, Northwestern University, 2145 Sheridan Road, Room C210, Evanston, IL, 60208, United States, David Morton We study a hierarchical clustering problem on an undirected graph with a weight function assigning nonnegative weights to the edges. We remove a subset of edges to break the graph into a number of smaller pieces, i.e., clusters. We consider a bicriteria graph clustering problem, in which we maximize the number of clusters while minimizing the weight of deleted edges. Solving this bicriteria problem parametrically identifies solutions that lie on the concave envelope of the efficient frontier, and the breakpoints on this envelope are nested, yielding a hierarchical family of clusters. We illustrate our ideas using NCAA football schedules, attempting to identify conferences, divisions, etc. 4 - Mixture-based Multiple Imputation Models for Clinical Data with a Temporal Dimension Ye Xue, Northwestern University, Evanston, IL, United States, Yuan Luo, Diego Klabjan Missing values are commonly occurred in many kinds of datasets, espectially in time series. We present mixture-based multiple imputation models for multi- variate time series. We design mixture models with one component capturing the correlation between variables and others catching the fluctuations of time series. The mixture models are optimized by maximazing their mixture likelihood functions using Estimation and Maximization (EM) algorithm. Then the imputation is performed based on the estimates of parameters of the optimized mixture models. We demonstrate the effectiveness of our imputation models on clinical data with a temporal dimension.

329