Seminar on Computer Science and Applied Mathematics

PROGRAM

UTORAK, 07.11.2017. u 14:15, Sala 301f, MI SANU, Kneza Mihaila 36
Tatjana Davidović, Matematički institut SANU
TIGHTER MODELS FOR BARGE CONTAINER SHIP ROUTING
This talk addresses the problem of optimal planning of a liner service for a barge container shipping company. Given estimated weekly demands between pairs of ports, our goal is to determine the subset of ports to be called and the amount of containers to be shipped between each pair of ports, so as to maximize the profit of the shipping company. In order to save possible leasing or storage costs of empty containers at the respective ports, our approach takes into account the repositioning of empty containers. The line has to follow the outbound-inbound principle, starting from the port at the river mouth. We propose a novel integrated approach in which the shipping company can simultaneously optimize the route (along with repositioning of empty containers), the choice of the final port, length of the turnaround time and the size of its fleet. To solve this problem, a new mixed integer programming model is proposed. On the publicly available set of benchmark instances for barge container routing, we demonstrate that this model provides very tight dual bounds and significantly outperforms the existing approaches from the literature for splittable demands. We also show how to further improve this model by projecting out arc variables for modeling the shipping of empty containers. Our numerical study indicates that the latter model improves the computing times for the challenging case of unsplittable demands. We also study the impact of the turnaround time optimization on the total profit of the company.
Joint work with L. Alfandari, F. Furini, I. Ljubić, V. Maraš, S. Martin.


UTORAK, 14.11.2017. u 14:15, Sala 301f, MI SANU, Kneza Mihaila 36
Marijana Petrović, Faculty of Transport and Traffic Engineering
AN ELECTRE-BASED DECISION AID TOOL FOR STEPWISE BENCHMARKING
Benchmarking is a management tool defined as a systematic process of measuring products, services or business practices against those recognized as business leaders. One way to adjust benchmarking to address successive performance improvements is found in the literature under the term stepwise benchmarking. The logic of stepwise approach is that a benchmarking unit should follow a sequence of relevant benchmarks along its improvement path toward target performance. An innovative approach for operationalization of stepwise benchmarking is using a specific ELECTRE-based procedure, labeled as ELECTRE Multi-Level Outranking. This outranking approach is used to establish the hierarchical position of benchmarking units. A specific procedure enables to select preferable development paths that consist of a sequence of intermediate benchmarks. The proposed method can help decision makers to 1) detect disparities among benchmarking units and 2) identify intermediate benchmarks for less successful ones in each stage of their progress toward the ultimate benchmark target.

Nataša Bojković, Faculty of Transport and Traffic Engineering
SUPPORTING PERFORMANCE APPRAISAL IN ELECTRE BASED STEPWISE BENCHMARKING MODEL
ELECTRE-based stepwise benchmarking model can be improved with data pre-processing procedure based on the decision maker's judgments about the performance levels. For this purpose, a new model translates the sequence of descriptive, non-numerical decision maker's assertions into scoring parameters. The proposed preference elicitation procedure enables to accommodate decision maker's perception on performance of benchmarking units. Thereby distinct performance characteristics are acknowledged and practical feasibility of benchmarking paths is enhanced. The important feature of the proposed procedure is that it allows proceeding with more scoring options, which can occur with some performance indicators.


UTORAK, 21.11.2017. u 14:15, Sala 301f, MI SANU, Kneza Mihaila 36
Milica Maričić i Nikola Zornić, Fakultet organizacionih nauka
MODEL RANGIRANJA FAKULTETA I INSTITUTA U OKVIRU UNIVERZITETA ZASNOVAN NA PERCENTILIMA
Metode rangiranja univerziteta vremenom su počele da dobijaju na značaju, a samim tim su i njihovi rezultati počeli da imaju sve veći uticaj na predstavnike vlasti i donosioce odluka. Međutim, tema koja još uvek ne privlači adekvatnu pažnju je rangiranje fakulteta i instituta u okviru univerziteta na osnovu naučnih rezultata. Cilj ovog predavanja je prikaz kreirane metode za rangiranje fakulteta i instituta na osnovu broja objavljenih radova, normalizovanog skora citiranosti (MNCS) i četiri indikatora bazirana na percentilima. Kao metod agregacije indikatora korišćeni su modeli multivarijacione analize. Kako bi se ispitala validnost predložene metode, ona je primenjena na institucije koje čine Univerzitet u Beogradu (UB). Rezultati pružaju značajne informacije o naučno-istraživačkim rezultatima 31 fakulteta i 11 instituta koji čine UB.


UTORAK, 28.11.2017. u 14:15, Sala 301f, MI SANU, Kneza Mihaila 36
Nenad Mladenović, Matematički institut SANU
DVADESET GODINA OD OBJAVLJIVANJA PRVOG RADA O METODI PROMENLJIVIH OKOLINA
Metoda promenljivih okolina (Variable Neighborhood Search, VNS) nastala je devedesetih godina prošlog veka u istraživačkom centru GERAD u Montrealu. Prvi rad u časopisu vezan za VNS je
N. Mladenović, P. Hansen, Variable neighborhood search, Computers and Operations Research 24(11) (1997), 1097–1100, doi:10.1016/s0305-0548(97)00031-2
i citiran je preko 3000 puta. Od tada je ova metoda uspešno primenjivana na mnoge probleme kombinatorne i kontinualne optimizacije. Brojne su i njene modifikacije, kao i hibridizacije sa drugim metodama. O razvoju i primenama VNS-a, kao i o istraživačima koji su u tome učestvovali, kako u svetu tako i u Srbiji, govoriće jedan od njenih kreatora.