Informs Annual Meeting 2017

TC71

INFORMS Houston – 2017

2 - Wildlife Corridor Sub-network Selection Optimizing for Probability of Connectivity Bistra Dilkina, Georgia Institute of Technology, 1304 Klaus Bldg, 266 Ferst Dr, Atlanta, GA, 30332, United States, bdilkina@gmail.com Corridor analysis results in a network representing the best pathways connecting pairs of habitat areas. One widely-used metric in ecology for the overall network connectivity is Probability of Connectivity (PC), where link probabilities are an exponential function of link resistance length. Due to limited budgets, the full corridor network usually cannot be conserved. Hence, we propose the PC-based Corridor Sub-network selection problem to select a subset of the corridor links of limited size or cost, while maximizing the PC of the protected part of the network. We present both a MIP model for finding optimal solutions of small networks, as well as a highly-scalable and effective greedy algorithm. 3 - Dynamic Optimization of Landscape Connectivity Embedding Spatial-capture-recapture Information Yexiang Xue, Cornell University, Institute for Computational Sustainability, 344 Gates Hall, Ithaca, NY, 14853-7501, United States, yx247@cornell.edu Maintaining landscape connectivity is increasingly important in wildlife conservation, especially for species experiencing the effects of habitat fragmentation. We propose a novel approach to dynamically optimize landscape connectivity. Our approach is based on a mixed integer program, embedding a spatial capture-recapture model that estimates the density, space usage, and landscape connectivity for a given species. Our method takes into account the fact that local animal density and connectivity change dynamically and non-linearly with different protection plans. We show that our method scales to real-world size problems and outperforms the solution quality of competing approaches. 371E Data Mining Contributed Session Chair: Misuk Lee, Seattle University, 901 12th Ave, Seattle, WA, 98122, United States, leem@seattleu.edu 1 - Development of a Dynamic Tool for Continuous Survival Analysis Transplant Survivor Hamidreza Ahady Dolatsara, Auburn University, 311 West Glenn The objective of this study is presenting a tool that can be utilized for data cleaning, feature extraction, and prediction on transplants’ survival analysis. Investigated transplants are kidney, heart, Lung, and intestine. Based on the dependent variable a dynamic process perform data cleaning process on the dataset. The cleaned data goes through feature selection and prediction phases. Well known machine learning algorithms are utilized in order to extract important features and develop predicting models. Finally, the output of models and other relevant information will be shown or emailed to the investigator(s). 2 - A Method to Measure Mobile Applications User Engagement (MAUE) Lior Turgeman, IBM.Research Laboratories, Haifa, Israel, tur.lior@gmail.com, Idan Ben Harush, Otis Smart, Max Jacubowsky, Nicole Jayne, Nili Ifergan We propose a new method to measure User engagement (UE) with mobile applications by analyzing temporal changes in a defined set of usage metrics, yielding a general metric - mobile applications user’s engagement (MAUE). Our proposed approach has been applied to the usage data of 40004 users of The Weather Company (TWC) application. Based on our results, we propose two new engagement measures, “engagement stability”, and “engagement breakpoint”, allowing monitoring UE of users’ segments, as well as gaining deeper understanding of the influence of different app updates, or certain features, on their usage behavior. 3 - A Customized Spectral Clustering Approach for Pre Selection on Dynamic Taxi Sharing Trips Yeming Hao, University of Maryland-College Park, College Park, MD, 20740, United States, yhao@umd.edu, Ali Haghani A customized spectral clustering approach is developed to narrow down the searching space for dynamic taxi sharing. Both taxi route geographical locations and heading directions are considered in the approach. Real world taxi data (New York City taxi data) is used in the implementation and results show the approach can reduce the taxi sharing matching calculation time while maintain beneficial (for both taxi providers and users) matching results. TC70 Ave., Apt. 28, Alabama, Auburn, AL, 36830, United States, hamid@auburn.edu, Fadel Mounir Megahed, Ali Dag, Ying-Ju Chen

4 - Modelling Passengers’ Airport Choice in Multi-airport Regions: A Data Mining Approach Misuk Lee, Assistant Professor, Seattle University, 901 12th Ave, Seattle, WA, 98122, United States, leem@seattleu.edu Air travelers’ airport choice is a critical component of transportation planning in many multi-airport regions. This study proposes the use of data mining models in passengers’ airport choice. We compare different data mining and statistical models using data collected in the Seoul/Incheon metropolitan area.

TC71

371F Applications of Global Optimization Sponsored: Optimization, Global Optimization Sponsored Session Chair: Ismet Sahin, Texas Southern University, 3100 Cleburne St, Houston, TX, 77004, United States, isahin@gmail.com 1 - Time-varying Linear and Semidefinite Programs Bachir El Khadir, Princeton University, Princeton, NJ, United States, bkhadir@princeton.edu, Amir Ali Ahmadi We study linear semidefinite programs whose data (e.g., the matrices A, b and c in the LP case) are not constant but vary polynomially with time. We show that, under some conditions, we can approximate the optimal value of these problems arbitrarily well by searching for solutions that are polynomial functions of time themselves. Furthermore, we show that the problem of finding the optimal polynomial solution of a given degree can be cast exactly as a semidefinite program. 2 - Hypergraph Theoretic Polyhedral Relaxation of a Class of 0-1 Polynomial Program Dongwoo Kang, Korea University, 145 Anam-Ro, Seongbuk-Gu, Seoul, 02841, Korea, Republic of, towarmer@korea.ac.kr, Hong Seo Ryoo Logical Analysis of Data (LAD) is a combinatorial optimization-based machine learning methodology. A key stage of LAD is pattern generation that can be cast as a 0-1 multilinear program in maximization form. We study its Boolean multilinear polytope via hypergraph theoretic analysis of the 0-1 representation of a set of data under analysis and discover a new class of strong(-er) valid inequalities, including facet defining inequalities. We demonstrate the usefulness of new results on benchmark data mining datasets. 3 - Accurate Beamforming by using Population Based Optimization Method Ismet Shin, Texas Southern University, Houston, TX, United States, isahin@gmail.com, Nuri Yilmazer In this paper, we have presented an accurate beamforming technique, which is a spatial filtering used in smart antenna technology, relies on population based optimization. The proposed algorithm uses a single snapshot of data taken from a Uniform Linear Array (ULA). Population based optimization algorithms are used to determine the optimal values of the complex valued antenna array weights. The complex antenna weight vectors adjust the phase shift at the output of each antenna element in a ULA which is analogous to the phase shifter in the analog system. The simulation results show that proposed algorithm accurately maximizes the signals in the direction of Signal of Interest and null Jammers. 4 - Electrical Power Grid Optimization using Semi-markov Decision Process (SMDP) Ismet Sahin, Texas Southern University, 3100 Cleburne Street, Houston, TX, United States, isahin@gmail.com, Abayomi Ajofoyinbo This paper presents a semi-Markov decision process (SMDP)- based model for electric power grid optimization in terms of available power and distribution configurations. For this purpose, a novel SMDP- based model is constructed for maximizing delivery configurations subject to power flowing through transmission line. Through optimal analysis, the conclusion is reached that the technique presented in this paper provides optimal performing configuration.

361