Publications with Fulltext
Permanent URI for this collectionhttps://hdl.handle.net/20.500.14288/6
Browse
13 results
Search Results
Publication Open Access AUC maximization in Bayesian hierarchical models(IOS Press, 2016) Department of Industrial Engineering; Gönen, Mehmet; Faculty Member; Department of Industrial Engineering; College of Engineering; 237468The area under the curve (AUC) measures such as the area under the receiver operating characteristics curve (AUROC) and the area under the precision-recall curve (AUPR) are known to be more appropriate than the error rate, especially, for imbalanced data sets. There are several algorithms to optimize AUC measures instead of minimizing the error rate. However, this idea has not been fully exploited in Bayesian hierarchical models owing to the difficulties in inference. Here, we formulate a general Bayesian inference framework, called Bayesian AUC Maximization (BAM), to integrate AUC maximization into Bayesian hierarchical models by borrowing the pairwise and listwise ranking ideas from the information retrieval literature. To showcase our BAM framework, we develop two Bayesian linear classifier variants for two ranking approaches and derive their variational inference procedures. We perform validation experiments on four biomedical data sets to demonstrate the better predictive performance of our framework over its error-minimizing counterpart in terms of average AUROC and AUPR values.Publication Open Access Pricing for delivery time flexibility(Elsevier, 2020) Savelsbergh, Martin; Department of Industrial Engineering; Yıldız, Barış; Faculty Member; Department of Industrial Engineering; College of Engineering; 258791We study a variant of the multi-period vehicle routing problem, in which a service provider offers a discount to customers in exchange for delivery flexibility. We establish theoretical properties and empirical insights regarding the intricate and complex relation between the benefit from additional delivery flexibility, the discounts offered to customers to gain additional delivery flexibility, and the likelihood of acceptance of discount offers by customers. Computational experiments, using an exact dynamic programming algorithm show that, depending on the setting, cost savings exceeding 30% can be achieved.Publication Open Access Inequity-averse shelter location for disaster preparedness(Taylor _ Francis, 2019) Gutjahr, Walter J.; Department of Industrial Engineering; Salman, Fatma Sibel; Hashemian, Mohammadmahdi; Faculty Member; Department of Industrial Engineering; College of Engineering; 178838; N/AWe study the problem of selecting a set of shelter locations in preparation for natural disasters. Shelters provide victims of a disaster both a safe place to stay and relief necessities such as food, water and medical support. Individuals from the affected population living in a set of population points go to, or are transported to the assigned open shelters. We aim to take both efficiency and inequity into account, thus we minimize a linear combination of: (i) the mean distance between opened shelter locations and the locations of the individuals assigned to them; and (ii) Gini's Mean Absolute Difference of these distances. We develop a stochastic programming model with a set of scenarios that consider uncertain demand and disruptions in the transportation network. A chance constraint is defined on the total cost of opening the shelters and their capacity expansion. In this stochastic context, a weighted mean of the so-called ex ante and ex post versions of the inequity-averse objective function under uncertainty is optimized. Since the model can be solved to optimality only for small instances, we develop a tailored Genetic Algorithm (GA) that utilizes a mixed-integer programming subproblem to solve this problem heuristically for larger instances. We compare the performance of the mathematical program and the GA via benchmark instances where the model can be solved to optimality or near optimality. It turns out that the GA yields small optimality gaps in much shorter time for these instances. We run the GA also on Istanbul data to drive insights to guide decision-makers for preparation.Publication Open Access Spatiotemporal prediction of infectious diseases using structured Gaussian processes with application to Crimean-Congo hemorrhagic fever(Public Library of Science, 2018) Şencan, İrfan; Torunoğlu, Mehmet Ali; N/A; Department of Industrial Engineering; Ak, Çiğdem; Ergönül, Önder; Faculty Member; Department of Industrial Engineering; Graduate School of Sciences and Engineering; School of Medicine; College of Engineering; N/A; 237468Background: Infectious diseases are one of the primary healthcare problems worldwide, leading to millions of deaths annually. To develop effective control and prevention strategies, we need reliable computational tools to understand disease dynamics and to predict future cases. These computational tools can be used by policy makers to make more informed decisions. Methodology/Principal findings: In this study, we developed a computational framework based on Gaussian processes to perform spatiotemporal prediction of infectious diseases and exploited the special structure of similarity matrices in our formulation to obtain a very efficient implementation. We then tested our framework on the problem of modeling Crimean-Congo hemorrhagic fever cases between years 2004 and 2015 in Turkey. Conclusions/Significance: We showed that our Gaussian process formulation obtained better results than two frequently used standard machine learning algorithms (i.e., random forests and boosted regression trees) under temporal, spatial, and spatiotemporal prediction scenarios. These results showed that our framework has the potential to make an important contribution to public health policy makers.Publication Open Access An improved lower bound on the competitive ratio of deterministic online algorithms for the multi-agent k-Canadian Traveler Problem(Finding Press, 2022) Shiri, Davood; Department of Industrial Engineering; Salman, Fatma Sibel; Faculty Member; Department of Industrial Engineering; College of Engineering; 178838We present an improved lower bound on the competitive ratio of deterministic online algorithms for the multi-agent k-Canadian Traveler Problem.Publication Open Access GoNDEF: an exact method to generate all non-dominated points of multi-objective mixed-integer linear programs(Springer, 2019) Department of Industrial Engineering; N/A; Türkay, Metin; Rasmi, Seyyed Amir Babak; Faculty Member; PhD Student; Department of Industrial Engineering; College of Engineering; Graduate School of Sciences and Engineering; 24956; N/AMost real-world problems involve multiple conflicting criteria. These problems are called multi-criteria/multi-objective optimization problems (MOOP). The main task in solving MOOPs is to find the non-dominated (ND) points in the objective space or efficient solutions in the decision space. A ND point is a point in the objective space with objective function values that cannot be improved without worsening another objective function. In this paper, we present a new method that generates the set of ND points for a multi-objective mixed-integer linear program (MOMILP). The Generator of ND and Efficient Frontier (GoNDEF) for MOMILPs finds that the ND points represented as points, line segments, and facets consist of every type of ND point. First, the GoNDEF sets integer variables to the values that result in ND points. Fixing integer variables to specific values results in a multi-objective linear program (MOLP). This MOLP has its own set of ND points. A subset of this set establishes a subset of the ND points set of the MOMILP. In this paper, we present an extensive theoretical analysis of the GoNDEF and illustrate its effectiveness on a set of instance problems.Publication Open Access Crowd-shipping service network design problem(Pamukkale University / Pamukkale Üniversitesi, 2022) Department of Industrial Engineering; Yıldız, Barış; Faculty Member; Department of Industrial Engineering; College of Engineering; 258791With the rising concerns over the increasing costs and negative externalities of city logistics, the need for innovative approaches to come up with low-cost and environmentally friendly distribution systems is on the rise. One such innovative approach is crowd-shipping (CS). In this model, which aims to combine packet and passenger transfers to utilize redundant transportation capacity, the individuals who want to earn small compensations and/or help the environment are employed to carry out package transfers between service points on the paths of their already planned trips. In this study, we introduce the strategic level CS network design problem, which needs to be solved to determine the locations and the capacities of the service points in the urban area to establish an efficient CS delivery system. We propose a novel Benders Decomposition algorithm to solve this challenging problem that we model as a scenario-based two-stage stochastic integer program. The results of our comprehensive numerical experiments attest to the efficacy of the approach to solve practical size problem instances and provide significant managerial insights, showing that such a well-designed CS network can provide savings (in terms of both economic and environmental costs), even in the case of express package transfers with very stringent delivery lead time restrictions. In particular, our results show that by carefully choosing the locations and the capacities of the service points, it is possible to carry out a significant portion of express deliveries with the crowd provided delivery capacity (up to 56% in our experiments) without deploying a very large number of them (less than 60 in all our experiment) in the region, indicating the strong potential and applicability of the CS delivery systems in real world delivery operations. / Şehir içi dağıtım faaliyetlerinin artan maliyetleri ve neden olduğu sorunlar (trafik, çevre kirliliği, gürültü, araç kazaları, vb.) hakkında artan kaygılar, bu alanda yenilikçi yaklaşımlara olan ihtiyacı giderek artırmaktadır. Ön plana çıkan söz konusu yaklaşımlardan birisi de Kitle destekli dağıtım (KDD) modelidir. Yolcu ve paket taşımacılığını bütünleştirerek atıl araç kapasitelerinin faydalı bir şekilde kullanılmasını amaçlayan bu modelde, kişilerin küçük ekonomik kazançlar sağlamak ve/veya dağıtım faaliyetlerinin çevreye verdiği zararları azaltılmasına katkı sunmak amacıyla kendi seyahatleri sırasında yolları üzerindeki servis noktaları arasında paket taşımacılığı yapması öngörülmektedir. Bu çalışmada, KDD faaliyetlerinin etkin bir şekilde yürütebilmesi için gerekli olan servis noktalarının/paket otomatlarının şehrin hangi noktalarına hangi kapasiteler ile kurulması gerektiğini belirlemek için çözülmesi gereken stratejik seviye “KDD servis ağı tasarımı” problemini literatüre tanıtıyoruz. İki seviyeli, senaryo tabanlı belirsiz bir tamsayılı program olarak modellediğimiz problemin çözümü için etkin bir Benders Ayrıştırma algoritması önermekteyiz. Gerçekleştirdiğimiz geniş kapsamlı hesapsal çalışmalar geliştirdiğimiz çözüm yönteminin etkinliğini ortaya koymakta, önemli yönetimsel çıkarımlar sunmaktadır. Elde ettiğimiz sonuçlar, hızlı paket taşımacılığı gibi oldukça zorlu ve maliyetli (ekonomik ve çevresel) bir dağıtım faaliyeti için iyi tasarlanmış bir KDD modelinin önemli kazanımlar sağlayabileceğini göstermektedir. Nispeten küçük sayıda (çalıştığımız problem örneklerinde en fazla 60 adet) servis noktaları ile dağıtımların önemli bir kısmının (çalıştığımız problem örneklerinde %56’ya kadar) “kitle” tarafından yapıldığı etkin bir KDD sisteminin teşkil edilebileceğini gösteren bu sonuçlar, önerilen yenilikçi yaklaşımın uygulanabilirliği konusunda önemli ipuçları vermektedir.Publication Open Access Integrating gene set analysis and nonlinear predictive modeling of disease phenotypes using a Bayesian multitask formulation(BioMed Central, 2016) Department of Industrial Engineering; Gönen, Mehmet; Faculty Member; Department of Industrial Engineering; College of Engineering; 237468Identifying molecular signatures of disease phenotypes is studied using two mainstream approaches: (i) Predictive modeling methods such as linear classification and regression algorithms are used to find signatures predictive of phenotypes from genomic data, which may not be robust due to limited sample size or highly correlated nature of genomic data. (ii) Gene set analysis methods are used to find gene sets on which phenotypes are linearly dependent by bringing prior biological knowledge into the analysis, which may not capture more complex nonlinear dependencies. Thus, formulating an integrated model of gene set analysis and nonlinear predictive modeling is of great practical importance. In this study, we propose a Bayesian binary classification framework to integrate gene set analysis and nonlinear predictive modeling. We then generalize this formulation to multitask learning setting to model multiple related datasets conjointly. Our main novelty is the probabilistic nonlinear formulation that enables us to robustly capture nonlinear dependencies between genomic data and phenotype even with small sample sizes. We demonstrate the performance of our algorithms using repeated random subsampling validation experiments on two cancer and two tuberculosis datasets by predicting important disease phenotypes from genome-wide gene expression data. We are able to obtain comparable or even better predictive performance than a baseline Bayesian nonlinear algorithm and to identify sparse sets of relevant genes and gene sets on all datasets. We also show that our multitask learning formulation enables us to further improve the generalization performance and to better understand biological processes behind disease phenotypes.Publication Open Access On the accuracy of uniform polyhedral approximations of the copositive cone(Taylor _ Francis, 2012) Department of Industrial Engineering; Yıldırım, Emre Alper; Faculty Member; Department of Industrial Engineering; College of EngineeringWe consider linear optimization problems over the cone of copositive matrices. Such conic optimization problems, called copositive programs, arise from the reformulation of a wide variety of difficult optimization problems. We propose a hierarchy of increasingly better outer polyhedral approximations to the copositive cone. We establish that the sequence of approximations is exact in the limit. By combining our outer polyhedral approximations with the inner polyhedral approximations due to de Klerk and Pasechnik [SIAM J. Optim. 12 (2002), pp. 875-892], we obtain a sequence of increasingly sharper lower and upper bounds on the optimal value of a copositive program. Under primal and dual regularity assumptions, we establish that both sequences converge to the optimal value. For standard quadratic optimization problems, we derive tight bounds on the gap between the upper and lower bounds. We provide closed-form expressions of the bounds for the maximum stable set problem. Our computational results shed light on the quality of the bounds on randomly generated instances.Publication Open Access Comparison of the results of blood glucose self-monitoring and continuous glucose monitoring in pregnant women with previous diabetes mellitus(Moscow Region Research and Clinical Institute (MONIKI), 2015) Dreval, A. V.; Shestakova, T. P.; Dreval, O. A.; Kulikov, D. A.; Medvedev, O. S.; Department of Industrial Engineering; Türkay, Metin; Faculty Member; Department of Industrial Engineering; College of Engineering; 24956Background: Pregnancy is one of the indications for continuous glucose monitoring (CGM). The data on its efficiency in pregnant women are contradictory. Aim: To compare the results of blood glucose self-monitoring (SMBG) and CGM in pregnant women with previous diabetes mellitus. Materials and methods: We performed a cross-sectional comparative study of glycemia in 18 pregnant women with previous type 1 (87.8% of patients) and type 2 diabetes (22.2% of patients) with various degrees of glycemic control. Their age was 27.7 ± 4.9 year. At study entry, the patients were at 17.2 ± 6.1 weeks of gestation. CGM and SMBG were performed in and by all patients for the duration of 5.4 ± 1.5 days. Depending on their hba1c levels, all patients were divided into two groups: group 1 – 12 women with the hba1c above the target (8.5 ± 1%), and group 2 – 6 women with the hba1c levels within the target (5.6 ± 0.3%). Results: According to SMBG results, women from group 2 had above-the-target glycemia levels before breakfast, at 1 hour after breakfast and at bedtime: 6.2 ± 1.6, 8.7 ± 2.1, and 5.7 ± 1.9 mmol/L, respectively. According to CGM, patients from group 1 had higher postprandial glycemia than those from group 2 (8.0 ± 2.1 and 6.9 ± 1.8 mmol/L, respectively, p = 0.03). The analysis of glycemia during the day time revealed significant difference between the groups only at 1 hour after dinner (7.1 ± 1.4 mmol/L in group 1 and 5.8 ± 0.9 mmol/L in group 2, р = 0.041) and the difference was close to significant before lunch (6.0 ± 2.2 mmol/L in group 1 and 4.8 ± 1.0 mmol/L in group 2, р = 0.053). Comparison of SMBG and CGM results demonstrated significant difference only at one timepoint (at 1 hour after lunch) and only in group 1: median glycemia was 7.4 [6.9; 8.1] mmol/L by SMBG and 6 [5.4; 6.6] mmol/L by CGM measurement (р = 0.001). Lower median values by CGM measurement could be explained by averaging of three successive measurements carried out in the period of rapid changes of glycemia. Conclusion: The achievement of control of diabetes by hba1c doesn't necessarily reflect current achievement of the target glycemic levels. As long as there was no significant difference in glycemia measured by SMBG and CGM, we conclude that CGM doesn't have any advantage over routine frequent SMBG in pregnant women.