Science

EURO XXIV

Stochastic Modelling and Simulation

Stream of European Conference on Operational Research

(July 11-14, 2010, Lisbon, Portugal)

Erik Kropat

University der Bundeswehr, Munich

Werner-Heisenberg-Weg 39, 85577 Neubiberg, Germany

Zeev Volkovich

ORT Braude Academic College

Yehiam 6, 21823, Karmiel, Israel

Gerhard-Wilhelm Weber

Middle East Technical University

Ankara, Turkey

EURO XXIV - the 24th European Conference on Operational Research was held from 11-14 July 2010 in Lisbon, Portugal. This conference was organized by the Association of European Operational Research Societies (EURO) within The International Federation of Operational Research Societies (IFORS). With more than 2700 participants and 700 sessions, this was the largest EURO conference ever. In the following we review the reports presented in the stream on "Stochastic Modelling and Simulation" organized by Erik Kropat (Universität der Bundeswehr München, Germany), Zeev Volkovich (ORT Braude College Karmiel, Israel) and Gerhard-Wilhelm Weber (Middle East Technical University, Turkey).

The stream "Stochastic Modelling and Simulation" consisted of the following 13 sessions:

-        New Achievements in Stochastic Models and Optimization

-        Stochastic Modelling and Simulation I, II

-        Markov Chains

-        Queueing Systems

-        Supply Chain Management

-        Simulation and Optimization of Networks under Uncertainty

-        Complex Systems under Uncertainty: Networks and Data Mining

-        Decision Analysis in Marketing and Financial Modelling

-        Data Mining and Knowledge Representation

-        Graph Theory and Combinatorial Optimization

-        Stochastic Models and Queueing Systems

-        Stochastic Models and Optimization.

Topics and Presentations

-        Markov chains;

-        Financial modelling;

-        Marketing;

-        Queueing systems;

-        Networks;

-        Optimization;

-        Supply Chain Management;

-        Decision analysis;

-        Data mining & Visualization;

-        Stochastic modelling & simulation.

Stochastic Modelling and Simulation

Theoretical and practical problems in the fields of statistical data analysis and visualization were addressed in the sessions "Data Mining and Knowledge Representation" and "Stochastic Modelling and Simulation I & II".

Presentations

"Fractal dimension cluster validation criteria".

Dvora Toledano-Kitai, Renata Avros, Zeev Barzily, Zeev Volkovich.

"Learning parameter optimization of stochastic gradient descent with momentum for a stochastic quadratic".

Memmedaga Memmedli, Engin Tas.

"A randomized algorithm for estimation number of clusters".

Zeev Volkovich, Oleg Granichin, Dmitry Shalymov, Renata Avros.

"On the number of coincidence of two homogeneous random walks with positive paces"

V.N.Surikov, I.A.Kravchenko.

"Comparing partitions: visual aids".

Ana Alexandra Martins, Margarida Cardoso.

"On the efficiency of spectral clustering: interpretation, parallel computation and results".

Sandrine Mouysset, Joseph Noailles, Danielle Ruiz.

"WhiBo - A platform for component-based design of partitioning cluster algorithms".

Kathrin Kirchner, Boris Delibasic, Milos Jovanovic, Milan Vukicevic, Johannes Ruhland.

In the presentation "Fractal dimension cluster validation criteria", Dvora Toledano-Kitai, Renata Avros, Zeev Barzily and Zeev Volkovich dealt with the cluster validation problem. Their approach is based on a fractal dimension cluster quality model. The cluster dimension values in a partition are repeatedly assessed by means of simulated samples occurrences in groups such that the proximity of the result obtained is interpreted as partition goodness. The empirical distribution most concentrated at the origin of the calculated dimension differences indicates the estimated number of clusters.

An optimization method particularly suited for classification problems with large and redundant data sets was the subject of the talk "Learning parameter optimization of stochastic gradient descent with momentum for a stochastic quadratic" of Memmedaga Memmedli and Engin Tas. The authors propose to learn the required parameters using the second-order information embedded in the Hessian. As a result, the convergence speed of stochastic gradient descent with adaptively tuned learning parameters can be compared with standard stochastic gradient descent on the Rosenbrock performance function.

The estimation of the number of clusters in a dataset was addressed by Zeev Volkovich, Oleg Granichin, Dmitry Shalymov and Renata Avros in their talk "A randomized algorithm for estimation number of clusters". In the proposed "elbow" method, the true number of clusters is recognized as the discontinuity point of the differential risk function, where the position is allocated by a randomized optimization algorithm. In particular, the numerical experiments demonstrate the good performance of the method and its low complexity cost in the case of a high number of clusters.

V.N.Surikov and I.A.Kravchenko gave a presentation with the title "On the number of coincidence of two homogeneous random walks with positive paces". Practical problems connected to genetic objects recognition and complex information systems monitoring often lead to objects identification tasks. The authors presented results that can be used to separation hypotheses about the membership of two samples to the same object in the presence of noisy data or to close different states of the same object.

In the presentation "Comparing partitions: visual aids", Ana Alexandra Martins and Margarida Cardoso addressed the problem of clustering evaluation and the visualization of the relationship between two partitions as well as the distance between the corresponding groups. In the proposed approach, the association between two partitions is depicted by a contingency table reflecting the similarities between two partitions. Multidimensional scaling is then applied for the visualization of the relationships between the partitions.

Sandrine Mouysset, Joseph Noailles and Danielle Ruiz gave new insights in spectral clustering in their talk "On the efficiency of spectral clustering: interpretation, parallel computation and results". In this approach, data points are clustered in a low dimensional data space by selecting dominant eigenvectors of a Gaussian affinity matrix. The authors reformulated the spectral clustering algorithm as an eigenvalues problem. In addition, a domain decomposition strategy for parallel spectral clustering was proposed.

In the presentation "WhiBo - A platform for component-based design of partitioning cluster algorithms", Kathrin Kirchner, Boris Delibasic, Milos Jovanovic, Milan Vukicevic and Johannes Ruhland proposed the design of new clustering algorithms by structuring and combining existing ones as sets of reusable components. For an evaluation the system was integrated in a white box machine learning platform for RapidMiner.

Queueing Systems

During EURO XXIV much attention was given to contributions to queueing systems. In the stream "Stochastic Modelling and Simulation", the sessions "Queueing Systems", "Stochastic Models and Queueing Systems" and "New Achievements in Stochastic Models and Optimization" were dedicated to this topic.

Presentations

"Analytical study of a queueing system with non-geometric tail behavior".

Mark van Lokeren, Bart Steyaert, Herwig Bruneel.

"Queueing system simulation and efficiency evaluation by Petri nets and data envelopment analysis".

Dragana Makajic-Nikolic, Gordana Savic, Mirko Vujosevic, Novak Novokmet.

"Queues with boundary assistance and the many effects of truncation".

Giang Nguyen, Peter Taylor, Guy Latouche.

"The effect of the number of servers in quasi-random-input queues".

Moshe Eben-Chaime.

"The effect of correlation in make-to-order systems".

Michael Zazanis.

"Equilibrium and socially optimal strategies in the M/G/1 queue with vacations".

Antonio Gomez-Corral.

Queueing systems with nongeometric tail behavior were analyzed by Mark van Lokeren, Bart Steyaert and Herwig Bruneel. In the system under consideration, two types of customers with different arriving behavior are involved. A customer of type A enters the system surely, whereas customers of type B show a so-called discouraged arrival, that means, the probability that an customer of type B enters the system depends on the total number of customers already present in the system. The authors discussed the non-geometric tail behavior of the total number of customers in the system and presented several performance measures.

Dragana Makajic-Nikolic, Gordana Savic, Mirko Vujosevic, Novak Novokmet presented their model of postal service queues with two channels based on Petri nets. Several scenarios for distinct rules of additional channel activation were simulated and data envelopment analysis was applied for efficiency evaluation.

The presentation of Giang Nguyen, Peter Taylor and Guy Latouche was concerned with queues with boundary assistance that can be modeled by a quasi-birth-and-death process. A simple example of such a queueing system is a model with two queues, where each queue has its own Poisson-stream of customers and its own server with an exponentially distributed service time. In this model, a free server can serve a neighboring customer, providing boundary assistance.

The effect of the number of servers in quasi-random input queues was addressed by Moshe Eben-Chaime. The finite source population of quasi-random input queues creates state dependent arrival rates. Since probabilities and performance measures admit no closed form expression, analytical results are difficult to obtain. The author presented new results and demonstrated the monotone effect of the number of servers on most performance measures.

The effect of correlations in the processing times of make-to-order systems was studied by Michael Zazanis. The analysis was based on Markovian queueing models and matrix-geometric techniques. In particular, correlated processes with exponential marginals derived from the Kibble-Moran-Downton bivariate exponential distribution were considered. In addition, assemble-to-order systems with correlated demands were analyzed.

Antonio Gomez-Corral studied the balking behavior of customers in the single-server queue with vacations. In the model under consideration, arriving customers can decide whether to enter the system or balk. This decision is based on a reward-cost structure that includes both the desire for service and the unwillingness to wait. The author identifies equilibrium strategies and socially optimal strategies.

Markov chains and decision analysis in marketing and financial modelling

Markov chains and decision support systems are useful tools for modelling in almost all fields of modern Operations Research. In the sessions "Markov Chains" and "New Achievements in Stochastic Models and Optimization" several authors presented their studies on both theoretical and practical aspects of these important statistical modelling approaches.

Presentations

"Modelling operational decisions in start-up firms".

Thomas Archibald, Kuangyi Liu.

"Hamiltonicity-trace conjecture for singularly perturbed Markov chains".

Vladimir Ejov, Nelly Litvak, Giang Nguyen, Peter Taylor.

"Prospective customer equity measurement and monitoring".

Nadine Losch, Klaus Möller.

"Strategic investment timing under asymmetric access charge regulation in telecommunications".

Takashi Shibata.

"On some antagonistic game related to majority voting".

Michael Khachay.

In the session on Markov Chains, Thomas Archibald and Kuangyi Liu presented a general Markov decision model for operations management of start-up firms addressing inventory, capacity expansion and marketing decisions. The profit maximizing objective is extended to a survival maximizing objective with a constraint on capital that can be more suitable for start-up firms.

In a presentation on theoretical aspects of Markov chains, Vladimir Ezhov and his co-authors Nelly Litvak, Giang Nguyen and Peter Taylor analyzed singularly perturbations of a Markov chain that correspond to a stochastic policy. They proved that the trace of the fundamental matrix is minimized at policies corresponding to Hamiltonian cycles.

Nadine Losch and Klaus Möller proposed a Markov chain approach to model the dynamics of customer equity. As a particular feature, a segment specific customer equity based on classification and regression trees was used.

Liberalized telecommunication markets were reviewed under the aspect of decision analysis in marketing and financial modelling by Takashi Shibata. The author investigated the effects of an asymmetric access charge regulation on competitive investment strategies. Since entrants of such a market have disadvantages at incumbents, an asymmetric access charge regulation for such asymmetric firms can stimulate competitive investment. In particular, it is shown that an entrant with a cost disadvantage has an incentive to invest as a leader under an asymmetric access charge regulation.

In the presentation "On some antagonistic game related to majority voting", Michael Khachay investigated the stability of collective decisions. The author gave the example of simple majority voting as a classical approach to aggregation of individual decisions suggested by a committee of experts. The stability of such a decision, subject to an exclusion of a fixed number of experts, was analyzed.

Supply Chain Management, Stochastic Models in Finance and Production

Supply chain management and stochastic models in finance are two of the most active fields in modern Operations Research and presentations on these topics are always an essential part of the EURO-conferences. Presentations related to these fields were given in the sessions "Supply Chain Management", "Stochastic Models and Optimization" and "New Achievements in Stochastic Models and Optimization".

Presentations

"Design and planning of green supply chains: A fuzzy approach".

Tânia Pinto-Varela, Augusto Novais.

"Interaction analysis of participants in supply chains".

Ivana Kovacevic, Biljana Panic.

"An application of the revised weighting method in vendor selection with price breaks".

Tunjo Perić, Zoran Babic.

"Operating overall production under chance constraints".

Baruch Keren, Zohar Laslo, Gregory Gurevich.

"Rollover optimization under uncertain regulatory approval date for products with bass demand rate".

Hiba El Khoury, Christian van Delft, Laoucine Kerbache.

"Inflation derivatives: HJM framework and market models".

Kwai-sun Leung, Lixin Wu.

As consumers become more and more aware of environmental issues, companies might think about a re-evaluation of their supply chain and the implementation of green policies. A restructuring of the manufacturing processes and an improved environmental performance might lead directly to financial gains. In addition, companies can find cost savings by reducing the environmental impact of their business processes. In the presentation "Design and planning of green supply chains: A fuzzy approach", Tânia Pinto-Varela and Augusto Novais addressed the optimal design and planning of logistic structures with regard to the trade-off between profit and environmental impacts. As a generic mathematical framework a so-called Resource-Task Network was proposed. For an analysis of the corresponding bi-level optimization a symmetric fuzzy linear programming approach was applied.

The role of the uncertain interaction and the impact of communication between the participants of a supply chain were discussed in the presentation "Interaction analysis of participants in supply chains" of Ivana Kovacevic and Biljana Panic. Transactional analysis was used to describe and reveal the behavioral patterns and social interactions of participants.

The problem of vendor selection and quantities supplied with price breaks was analyzed in the presentation "An application of the revised weighting method in vendor selection with price breaks" of Tunjo Perić and Zoran Babic. The authors investigated the possibility of an application of the revised weighting method and tested the proposed method on the example of flour purchase.

The manufacturing process of a single product in a certain number of plants with heterogeneous characteristics and specific stochastic production capabilities was analyzed by Baruch Keren, Zohar Laslo, Gregory Gurevich in the presentation "Operating overall production under chance constraints". The authors presented a model that can be used to determine the required budget and its distribution among the plants in order to ensure the fulfillment of the periodic demands of the product according to due dates and pre-given confidence levels.

In the presentation "Rollover optimization under uncertain regulatory approval date for products with bass demand rate", Hiba El Khoury, Christian van Delft and Laoucine Kerbache were concerned with the decision problem of a company that plans to phase-in a new product and to phase-out an existing one. The production of the existing product should be neither stopped to early nor to late, because of potential losses in profit or unnecessary costs. The authors studied the rollover of products that follow a Bass demand rate, with the new product subject to an uncertain regulatory approval date. In particular, the costs associated with the rollover are minimized and optimal rollover strategies are determined.

Kwai-sun Leung and Lixin Wu established in their talk "Inflation derivatives: HJM framework and market models" a Heath-Jarrow-Morton type framework that governs the co-evolution of the term structure of both nominal and inflation rates. Based on this framework, the authors proposed a market model with forward inflation rates using displaced diffusion processes. This resulted in a closed-form pricing for inflation caplets and inflation swaptions.

Complex Systems and Networks under Uncertainty

One particular focus of the stream on "Stochastic Modelling and Simulation" was set on OR-Methods and applications for complex networks under uncertainty. Various topics related to simulation, optimization, forecasting and data mining were addressed in the sessions "Simulation and Optimization of Networks under Uncertainty", "Complex Systems under Uncertainty: Networks and Data Mining" and "Graph Theory and Combinatorial Optimization".

Presentations

"Extinction probabilities of decomposable branching processes".

Sophie Hautphenne.

"Search of the best alliance composition in an interdomain network".

Daniel Villa Monteiro, Thierry Mautor, Dominique Barth.

"Quantitative release planning in extreme programming".

Tommi Tervonen, Gert van Valkenhoef, Bert de Brock, Douwe Postmus.

"Synergic process of speech signal energy transmission".

Vladimir Zhuravlev, Dorovshykh Anatoliy.

The presentation of Sophie Hautphenne was concerned with extinction probabilities of decomposable branching processes. Such processes arise for example in telecommunication and biological systems when some types of individuals are not able to generate other types. In this situation, partial extinction of some classes can occur whereas the whole process may not necessarily become extinct. The authors established criteria for partial as well as total extinction and analyzed in detail the partial (total) extinction probability.

Daniel Villa Monteiro and his co-authors Thierry Mautor and Dominique Barth addressed the problem on how to find an optimal alliance composition in an interdomain network. In this particular network, the domains have to satisfy service requests. A subset of domains, where members share knowledge and provide a routing service to other members is considered as an alliance. Both heuristics and exact solution methods have been applied to find the best alliance that gives to the members the highest increase in service satisfaction.

Tommi Tervonen and his co-authors Gert van Valkenhoef, Bert de Brock and Douwe Postmus addressed "Quantitative release planning in extreme programming". This methodology tries to improve software quality with regard to changing customer requirements and is considered as a particular agile software development methodology with frequent releases in short development cycles. The authors provided a multiple knapsack model to assist release planning where the selected plan maximizes the expected business value.

In the presentation "Synergic process of speech signal energy transmission", Vladimir Zhuravlev and Dorovshykh Anatoliy analyzed the imperfect adequacy of speech and hearing process theories. They propose a mathematical model for speech signal energy generation and receiving that is based on synergic analysis of speech signal informational components in communication channel energy transfer speed and substance carriers.

Assignment Problems, Graph Theory and Combinatorial Optimization

A variety of theoretical contributions ranging from assignment problems and graph-theoretical models to combinatorial spaces were presented in the sessions on "Complex Systems under Uncertainty: Networks and Data Mining" and "Stochastic Modelling and Simulation I &II".

Presentations

Pavlo Krokhmal.

"Probabilistic analysis of multidimensional assignment problems".

"Approximation algorithms for the general multi assignment problem".

Ron Adany, Sarit Kraus, Fernando Ordonez.

"On the cubical dimension of four new classes of trees".

Kamal Kabyl, Abdelhafid Berrachedi.

"On defining a combinatorial space"

Sergii Sirenko.

"On goodness-of-fit tests for random combinatorial objects".

Alexander Kolodzey.

"Identification of local distortions in random sequences".

Alexander Grusho, Elena Timonina, Zeev Volkovich.

In the presentation "Probabilistic analysis of multidimensional assignment problems" Pavlo Krokhmal investigated a generalization of linear and quadratic assignment problems. Properties of large-scale randomized instances of multidimensional assignment problems under the assumption that their assignment costs are independent and identically distributed random variables are considered.

Multidimensional assignment problems were also addressed in the presentation "Approximation algorithms for the general multi assignment problem" of Ron Adany, Sarit Kraus and Fernando Ordonez. The authors presented a problem where personalized advertisements have to be assigned to viewers in order to maximize revenue. It is assumed that each viewer has a limited capacity and each advertisement has a given length whereas a revenue is obtained if is assigned to a certain number of viewers. Two bi-criteria approximation algorithms of the Ads Packing Problem were presented: Extra-Packing and Deep-Search-Replacer.

In a theoretical contribution, Kamal Kabyl and his co-author Abdelhafid Berrachedi considered the embedding of trees in a hypercube. The major goal is to find the smallest dimension of a hypercube in which a given tree is embeddable. In particular, the authors identified four new classes of trees for which the cubical dimension is established.

In another presentation on theoretical aspects of combinatorial optimization, Sergii Sirenko introduced a new formal approach for defining important notions such as a combinatorial optimization problem, a combinatorial space and combinatorial objects. With this new perspective countable combinatorial spaces can be considered. In addition, with the proposed definition of directed segments it is possible to provide constructive mechanisms for solving of combinatorial optimization problems.

Alexander Kolodzey gave a presentation on "On goodness-of-fit tests for random combinatorial objects". The author studied random combinatorial objects which possess the decomposition property of the individual components. The weight of the objects equals the sum of the components weights. In addition, a joint distribution of the number of components with a given weight can be represented as the joint conditional distribution of several independent Poisson random variables.

In the talk "Identification of local distortions in random sequences", Alexander Grusho, Elena Timonina and Zeev Volkovich consider sequences of consistent random distributions. As the main result, the authors demonstrate that it is possible to find local distortions in a random sequence, where the corresponding distribution is well-defined.

Conclusion

With our stream "Stochastic Modelling and Simulation" at EURO 2010 we contributed to a vibrant field of modern OR that is characterized by its connections to all the other areas of OR, in theory, methods and applications. Herewith, we organizers hope to having contributed to a closer exchange inside of the OR communities and between its scientists from all over the world. We are thankful to EURO, the EURO Working Groups and to IFORS, for their continuous support, and we would be glad if our stream became and will become a precious contribution to academia and to overcoming of the OR community's and humankind's future challenges.

Related links

Website of 24th European Conference on Operational Research (EURO XXIV)

www.euro2010lisbon.org

EURO - The Association of European Operational Research Societies within IFORS

www.euro-online.org

IFORS - The International Federation of Operational Research Societies   www.ifors.org




[Contents]

homeKazanUniversitywhat's newsearchlevel upfeedback

© 1995-2008 Kazan State University