ὅδε οἶκος, ὦ ἑταῖρε, μνημεῖον ἐστιν ζῴων τῶν σοφῶν ἀνδρῶν, καὶ τῶν ἔργων αὐτῶν

Seminar for
DECISION MAKING – THEORY, TECHNOLOGY AND PRACTICE

 

PROGRAM


Plan rada Seminara Odlučivanje - teorija, tehnologija, praksa za NOVEMBAR 2022.




ČETVRTAK, 03.11.2022. u 13:00, Live stream Niš
Željko Stević, University of East Sarajevo, Faculty of Transport and Traffic Engineering Doboj
OBJECTIVE CRITICISM AND NEGATIVE CONCLUSIONS ON USING THE FUZZY SWARA METHOD IN MULTI-CRITERIA DECISION MAKING
The quality of output or decision-making depends on high-quality input data, their adequate evaluation, the application of adequate approaches, and accurate calculation. In this paper, an objective criticism of applying the fuzzy SWARA (step-wise weight assessment ratio analysis) method based on the Chang TFN (triangular fuzzy number) scale is performed. Through research, it has been noticed that a large number of studies use this approach and, as an epilogue, there are wrong decisions based on inconsistent values in relation to the initial assessment of decision-makers (DMs). Seven representative studies (logistics, construction industry, financial performance management, and supply chain) with different parameter structures and decision matrix sizes have been singled out. The main hypothesis has been set, which implies that the application of this approach leads to wrong decisions because the weight values of the criteria are incorrect. A comparative analysis with the improved fuzzy SWARA (IMF SWARA) method has been created and a number of negative conclusions has been reached on using the fuzzy SWARA method and the Chang scale: Primarily, that using such an approach is impossible for two or more criteria to have equal value, that allocating TFN (1,1,1) leads to criteria values that are inconsistent with expert evaluation, that the last-ranked criteria in the fuzzy SWARA method have no influential value on the ranking of alternatives, that there is a great gap between the most significant and last-ranked criteria, and that the most significant criterion has a huge impact on the evaluation of alternative solutions and decision making. As a general conclusion, it is given that this approach is not adequate for application in problems of multi-criteria decision making because it produces inadequate management of processes and activities in various spheres.



ČETVRTAK, 10.11.2022. u 13:00, Live stream Niš
Anđelka Zečević, Mathematical Institute SANU
EVALUATION OF DOCUMENT SUMMARIES
Natural Language Processing is a multidisciplinary field that aims to understand and model the underlying principles of human language. One of the tasks within the area that is particularly interesting these days due to the overall information overload is document summarization which exploits the techniques of generating short and relevant summaries for a given document. In this talk, we will cover challenges met by evaluating state-of-the-art summarization models and specifically address the problem of information diversification.

ČETVRTAK, 24.11.2022. u 13:00, Live stream Niš
Dragan Urošević, Mathematical Institute SANU
GENERAL VARIABLE NEIGHBORHOOD SEARCH FOR THE MULTIPLE TRAVELING REPAIRMAN PROBLEM WITH PROFITS
There are set of n users (labeled by numbers between 1 and n) requiring some of kind of service and set of m (identical) repairmans (numbered by numbers between 1 and m). Each repairman starts from initial position (depot), visits and services users in some order. The time of servicing of each of users is equal to zero. For performed service, repairman obtains revenue which is equal to difference between initial profit proposed by user (pi for user i) and time until this user is serviced. The problem consists in selecting a subset of all users (customers) to be served by some of these repairmans so that the total revenue received from the users (customers) is maximized. To address this problem, an general variable neighborhood search is proposed. The main part of proposed method is variable neighborhood descent as local search. Proposed variable neighborhood search consists of exploring in total e even different neighborhoods. For representing solution an efficient structure (the Frenwick tree) is used in order to speedup neighborhoods exploring. The performance of the algorithm were compared with those of the leading reference algorithms on selected subset of 470 instances.


Predavanja su namenjena sirokom krugu slusalaca, ukljucujuci studente redovnih i doktorskih studija. Seminar će se održavati svakog drugog četvrtka od 13:00 - 14:00h, CIITLAB, Elektronski fakultet Niš, Aleksandra Medvedeva 14, Niš

dr Lazar Velimirović
Rukovodilac seminara
dr Petar Vranić
Sekretar seminara