Milija Suknović’s research while affiliated with University of Belgrade and other places

Introduction/purpose: Wheeled armored combat vehicles are combat systems that are increasingly present in modern armed conflicts, especially in operations against asymmetric threats. The global wheeled armored vehicle market is constantly growing, which reflects their application in a wide range of missions and tasks of armed forces. The existence of numerous models of these vehicles with different technical and exploitation characteristics, along with the possibility of adaptation to specific needs, further complicates the choice of the most suitable alternative. The paper presents the case of solving the problem of selecting the most suitable multi-purpose medium-class wheeled armored vehicle with a 4x4 drive formula when choosing one of the four alternatives, using individual and group multi-criteria decision-making methods. Methods: In the paper, the methods of multi-criteria decision making were applied to solve problems in the field of complex combat systems selection. Experts from the field of tactics with weapon systems have ranked the alternatives following the defined criteria using the AHP (Analytic Hierarchy Process) and the PROMETHEE II (Preference Ranking Organization Method for Enrichment Evaluations II) methods. The results obtained by individual decision making were subjected to the Condorcet method of group decision making to make a final decision. Result: Selection of the most suitable vehicle by the defined criteria. Conclusion: Solving the problem involves taking into account the views of military experts regarding the optimization of multiple criteria to provide the best performance vehicle suitable for use in various missions. The choice of a multi-purpose wheeled armored combat vehicle is a complex process influenced by numerous factors that cannot be analyzed objectively without the application of adequate mathematical models.

Nowadays, decision-making systems in modern infrastructural planning greatly impact everyday life. This paper proposes a novel modification of the multi-criteria decision analysis (MCDA) method VIKOR that can be successfully applied to infrastructural decision-making systems. Our contributions are twofold: We first solve a highway section selection on the Montenegro A1 highway. Secondly, we modify the VIKOR method for the multi-actor (MA) setting. Although the original VIKOR method recognized multi-actor preferences through the selection of the value of the compromise parameter v, it did not explicitly include multiple actors in the decision-making process. Moreover, we show how the multi-actor (MA) VIKOR method can serve as a decision support system for making important infra-structural decision problems, improve the transparency of the decision-making process with the rising need to include citizens in the decision-making process, and how it successfully solves the distortion in social choice problem.KeywordsMCDAVIKORhighway selectionmulti-actor

Using machine learning algorithms in social environments and systems requires stricter and more detailed control. More specifically, the cost of error in such systems is much higher. Therefore, one should ensure that important decisions, such as whether to convict a person or not based on the previous criminal record, are by the legal requirements and not biased toward a group of people. One can find many many papers in the literature aimed at mitigating or eliminating unwanted bias in machine learning models. A significant part of these efforts add fairness constraint to the mathematical model or adds a regularization term to the loss function. In this paper, we show that optimizing the loss function given the fairness constraint or regularization for unfairness can surprisingly yield unfair solutions. This is due to the linear relaxation of the fairness function. By analyzing the gap between the true value of fairness and the one obtained using linear relaxation, we found that the gap can be as high as around 21% for the COMPAS dataset, and around 35% for the Adult dataset. In addition, we show that the fairness gap is consistent regardless of the strength of the fairness constraint or regularization.

This paper addresses a novel topic in ski injury research - how a ski group size indicates the risk of ski injury. There is evidence in research literature that people ski in groups. However, the relationship between group size and the risk of injury has remained unexplored. Based on ski lift entrance data, we use the density-based clustering algorithm OPTICS to identify groups of skiers and discuss the advantages of using this algorithm. We show that the ski group size can be used to improve the identification of skiers who experience ski injury. The results of the identification of ski groups at Mt. Kopaonik Ski Resort in Serbia show that skiing alone is most susceptible to ski injury, while skiing in couples or in bigger groups reduces the risk of injury by 46%. In addition, it is confirmed that ski injuries are an early failure event phenomena. Based on the CHAID decision tree analysis, spending a small amount of time at the ski resort and skiing alone are associated with the: 6 times greater ski injury risk than the average ski injury risk.

Achieving fairness in algorithmic decision-making tools is an issue constantly gaining in need and popularity. Today, unfair decisions made by such tools can even be subject to legal consequences. We propose a new constraint that integrates fairness into data envelopment analysis (DEA). This allows the calculation of relative efficiency scores of decision-making units (DMUs) with fairness included. The proposed fairness constraint restricts disparate impact to occur in efficiency scores, and enables the creation of a single data envelopment analysis for both privileged and unprivileged groups of DMUs simultaneously. We show that the proposed method - FairDEA - produces an interpretable model that was tested on a synthetic dataset and two real-world examples, namely the ranking between hybrid and conventional car designs, and the Latin American and Caribbean economies. We provide the interpretation of the FairDEA method by comparing it to the basic DEA and the balanced fairness and efficiency method (BFE DEA). Along with calculating the disparate impact of the model, we performed a Wilcoxon rank-sum test to inspect for fairness in rankings. The results show that the FairDEA method achieves similar efficiency scores as other methods, but without disparate impact. Statistical analysis indicates that the differences in ranking between the groups are not statistically different, which means that the ranking is fair. This method contributes both to the development of data envelopment analysis, and the inclusion of fairness in efficiency analysis.

Recently, the usage of machine learning algorithms is subject to discussion from a legal and ethical point of view. Unwanted discrimination regarding gender or race of a prediction model can lead to legal consequences. Therefore, during predictive model learning, one needs to be aware of possible bias and adjust the model to be fair. However, in bank marketing applications, one customer can receive multiple offers instead of just one. Because of their correlation between, a multi-label classification approach is the most suitable one. This paper proposes a fair classifier chain machine learning model for multi-label classification. Our algorithm solves the multi-label classification problem in an efficient manner, and it is suitable for real-life application employment. The proposed approach allows for controlling fairness constraints during the process of machine learning. It is based on the logistic regression model, thus enabling high efficiency and understandability. We apply our model to a real-life model from bank marketing campaign response prediction. The obtained results are promising. More specifically, our model achieves high fairness measures having an increase from 7% to 17%. However, fairness has a price of a decrease in predictive performance, up to 9% of AUC. To the best of our knowledge, this is the first algorithm that introduce fairness constraints in multi-label classification problems.