Abstract: In this paper we examine the scientific production of Serbian scientists in the area of computing, published between year 2008 and year 2020, extracted from the Web of Science (WoS) database. Our dataset includes papers published in journals with impact factor, which are classified in at least one out of seven WoS categories with the prefix "Computer Science". The work includes identification of unique authors' and institutions' identity from the dataset, as well as creation of a database used for bibliometric and coauthorship analysis. As a result, we have created coauthorship network using Gephi, identified the leading authors and institutions, as well as the most popular journals, WoS categories and collaborators of Serbian authors in the field of computer science for the analyzed period.

Abstract: Compliant grippers are used in the macro and micro world for capturing and manipulating objects, when it is necessary to manipulate fragile objects, such as surgical instruments, in biomedicine for manipulating individual cells, etc. In addition, grippers with a rigid contact between the grippers and the rigid objects can lead to damage to the objects, so in these cases, a control system is required to control the contact forces. Development of compliant finger-like gripper for soft touch application is described in this paper. This paper provides an answer to the question how to design, model and implement such compliant gripper. The developed compliant fingers are inspired by the real human fingers. Machine learning methods where proposed for realization of non-distractive grasping of fragile objects where feedback information is acquired by computer vision system while intelligent algorithms where considered for motion control and grasping. The complexity of lays not only in non-destructive grasping but also in adequate force implementation in order to avoid dropping the grasped object. Hence, nonlinear simulation of compliant gripper, by means of finite elements method (FEM), is provided in this paper. Several experiment where done to provide real life conformation of such analysis.

Abstract: This research aims to propose a novel framework for the assessment of the consequences of hazardous events on a water resources system using dynamic resilience. The two main types of hazardous events considered are: a severe flood event, and an earthquake. Given that one or both hazards occur, this framework utilizes a digital twin based on a system dynamics (SD) model, backed by an Artificial Neural Network (ANN) to estimate the dynamic resilience. The ANN was trained using a large, simulated dataset ranging from very mild to extreme hazard combinations. The ANN’s efficacy was quantified using the average relative error metric which equals 2.14% and 1.77% for robustness and rapidity, respectively.

Abstract: In this paper, we will present novel general model for the irradiance FSO fluctuations, based on the Chi-square distribution. The error performance of the Free Space Optical (FSO) system modulated with On-Off keying (OOK) scheme both in the presence of atmospheric turbulence and misalignment fading (i.e. pointing error) will be investigated. For both cases, the expressions for the Average Bit Error Rate (ABER) will be determined in analytically closed form and discussed in function of average optical power at the transmission. The results will be graphically presented in order to determine impact of relevant parameters on the quality of the received signal in the OOK modulated FSO system.

Abstract: Data protection is given as an interlinked process of monitoring sensitive data within the frameworks, regulations and jurisdiction of a country, region or specific organization. Where each country based on the GDPR appoints a data protection authority which is responsible for investigating cybercrimes, correcting data and giving advice on data protection. The cyber threats developed in the Western Balkans threatened privacy and data protection in almost every country of this region, where many sensitive data fell into the hands of hackers or many systems were blocked for a certain period of time. To prevent these two types of problems, it is necessary to assign controls or filters to reduce the risk of unauthorized access to sensitive data. In our research, the collected data is classified into three main categories (private, public, private), which are then analyzed based on the amount and type of data circulating in the implemented systems, the type of encryption used, encryption configurations and data backup, a solution is provided in data protection by defining access controls to sensitive data. The purpose of our research is to provide insight into the current state of data protection in governmental, public institutions, and various service companies in order to provide a new solution that offers data protection which is based on the specifics of the amount and type of data, encryption and coding of data before and during transmission and data backup.

Abstract: As many industries are in high demand for Machine Learning (ML) practitioners to solve business problems, it is essential to ensure that students know how to select adequate ML tools for given contexts and apply them adequately. To this aim, we designed a project-based undergraduate university ML course. The course utilizes a blended approach, in which students collaboratively work on real-world projects using an autograding platform for code-based assignments specially developed for the needs of the course. The course includes traditional lectures, discussions, reporting, and oral presentations. The course was evaluated using class assessment outcomes, faculty surveys, and observations. The results indicated that the blended learning approach was wellreceived and helped students better understand how to apply ML tools. They also suggest that project-based learning, in combination with an autograding platform and a blended approach, can be an effective way to teach undergraduate ML.

Abstract: The graph coloring problem has a wide range of applications in different areas of science and engineering. Graph coloring is a problem where it is necessary to assign colors to graph elements (vertices or edges). The graph coloring problem can be represented in the way that it is necessary to color the vertices of the graph, or to color the edges of the graph under certain conditions. It belongs to the Non polynomial (NP) class of problems. This paper presents an analysis of algorithms for solving the edge coloring problem and their performance for bipartite graphs. For edge coloring, it is necessary to examine whether it is possible to color the edges of the graph with certain number of colors so that any two edges incident to the same vertex are colored with a different color. Numerous approaches have been developed to solve this problem, which have advantages and disadvantages. In this study algorithms for edge coloring were analyzed, such as brute force algorithm based on the maximum degree of the nodes, an algorithm that uses Dinitz’s method, and an algorithm that uses Euler’s partition. The solutions were carefully evaluated using four different datasets. The results show that brute force algorithm is superior for smaller graphs with up to 104 vertices, when it reaches memory limits. The algorithm that uses the Euler’s partition is the fastest for graphs with large number of vertices

Abstract: In this paper, the fluid flow between two parallel plates is considered using the SIMPLE method in the MATLAB software package. Velocity changes as a function of time is examined. The results obtained numerically coincided with the results given by the SIMPLE method. The ANSYS software package, module CFX was used for numerical proof. The acronym SIMPLE stands for Semi-Implicit Method for Pressure Linked Equations. The algorithm was originally put forward by Patankar and Spalding and is essentially a guess and-correct procedure for the calculation of pressure on the staggered grid. The method is illustrated by considering the two-dimensional laminar steady flow equations in Cartesian coordinates

Abstract: A lack of security controls in the system may be a potential point of attack to exploit the system's vulnerability. Unfortunately, security controls are added as an afterthought when all functionalities are implemented, which leads to difficulties adapting to the software's rigid policies, decreased performance, and increased costs. Furthermore, even after adjusting those policies, using an application that only partially fulfills some desired security requirements is difficult. The solution for decreasing time on adapting security mechanisms and minimizing weak points in the system becomes integrating security as a building block of development and maintenance, known as the DevSecOps concept. In this paper, we illustrate the importance of continuously providing protection in containers and reducing the risk of unwanted application attacks by integrating a secure pipeline in the earliest stage. The proposal is to automate the pipeline by combining security tools with database knowledge in a development process. The database knowledge will provide security policies that can be applied to a specific pipeline stage. This paper presents an approach to minimize vulnerabilities and code flaws by practicing DevSecOps, which also requires collaboration and communication between development, security, and operations teams, which increases the software's overall development efficiency.

Abstract: With the rapid development of technology, the demand for mobile devices is increasing more than laptops, because their main features are large memory, great cameras, low price, and easy availability. Today, we keep too much information on our devices which is very important for our life with this there is possibility that our device has been stolen. A mobile device is very important in investigation making mobile forensics very important for court proceedings and criminal investigations. The paper highlights the importance of specialized forensic tools in mobile forensics, catering to different operating systems (Android and iOS) and offering a combination of free and paid options. These tools serve various functions such as data acquisition, examination, recovery, and extraction. The paper discusses and present a selection of free and paid tools used in mobile forensics, categorizing them based on operating system compatibility, cost, and functions. It provides a summary of each tool's features and their significance in forensic investigations. The paper also concludes by emphasizing the need for professionals in the field to stay updated with the latest advancements to enhance their capabilities in uncovering crucial information, protecting data, and maintaining the integrity of digital investigations.

Abstract: One of the main challenges in primary food production is maintenance of the crops, and assuring optimal yield. However, very often, due to various reasons, plant density can be reduced which leads to reduced yield and longevity of the crops. First step in solving this problem is to address assessment of the actual plant density, which can be automated using images obtained via UAVs, and artificial intelligence methods. In this paper we present software solution for plant density assessment on agricultural land using neural networks on images obtained from UAVs. The proposed software solution covers the entire workflow, starting from training the model on the training set of images, and ending with the use of the trained model. This solution will provide support in decision making regarding condition of the plantations and further agrotechnical measures.

Abstract: IoT fitness trackers are widely used to measure vital parameters and track sports activities of their users. Available options allow uploading data from an IoT device through a mobile application to its cloud storage. There is a limitation that only data from users with the same devices or operating systems can be collected on the same cloud. The main goal of this work is to propose a solution for tracking and collecting data from heterogeneous IoT devices for fitness tracking and storing it in a shared database. This solution is intended for use in fitness clubs where members, users of IoT fitness trackers, can share their data in the club's private cloud with regulated rights and privileges. This would enable professional fitness trainers to effectively track fitness activities and vital parameters, monitor progress, and provide individualized recommendations. Certain aspects of this solution have been fully developed and implemented, while others are presented as challenges for future resolution and improvement. The research and implemented activities, as well as the insights obtained, show that a recommended software technology stack can be based on implementing common functionalities and addressing the heterogeneity gap of devices, operating systems, working platforms, applications, and datasets. The functionality of this recommended technology stack is demonstrated with a concrete example, and its general applicability is also explained. One of the challenges addressed in this paper is how to integrate the proposed software technologies into a cohesive and functional system.

Abstract: Static and dynamic analysis of complex geometry is important to both the scientific community and industry. Based on various types of materials and loads, the stress and strain field on geometry could be different. Prediction of the stress and strain field using the input parameters, which define the geometry and material, can have a great impact on static and dynamic behavior. In this paper structural analysis of cracked and non-cracked beam is investigated. The bending analysis of the cantilever beam is considered by varying different parameters, such as length and height of the beam, force magnitude, number and position of the cracks, different types of materials etc. The results of stress and strain are obtained by FEM analysis and used as output for machine learning (ML) algorithm. Decision Tree (DT) and Random Forest (RF) are used as ML algorithms in which RF is evolved from the decision tree used as an ensemble approach, whereby the main goal is to use collection of trees for higher accuracy. Evaluation of the trained model has been done by mean absolute error and mean squared error. The workflow is divided into five steps: (a) acquisition of the data by FEM analysis, (b) feature engineering and data preprocessing, (c) training of ML algorithms, (d) evaluation and metric of trained model, (e) results and discussion.

Abstract: While the ethical principles of finance are well known in the literature, they are not sufficiently evaluated in the context of machine learning (ML). We use natural language processing (NLP) transformer models to quantitatively evaluate the relationships between the ethical principles of finance and the ethical principles of ML. To the best of our knowledge, such analysis has not been performed in the literature. We assess the performance of more than 80 state-of-the-art (SOTA) transformer models in capturing semantic similarity between the definitions of finance and ML ethics principles. The computational results demonstrate the ability of various transformers to address semantic similarity when comparing the definitions of finance and ML ethics. The results reveal that the NLI-DistilRoBERTa-Base-v2 model has the best performance in this task. The analysis can be beneficial to identify the principles of finance ethics that exhibit the strongest influence on ML ethics and vice-versa.

Abstract: Customer retention is one of the primary pillars of product growth with the subscription-based business model. The competition is fierce in the SaaS (Software as a service) market, where customers are free to choose from the many companies that offer similar competitive services. Sometimes, multiple bad experiences, or even a single one, can make the customer give up on the product or on the service. Globalization and rising competition are increasing the cost of getting a new customer, making it more affordable to invest in retaining customers. Therefore, it is crucial for any business to be able to anticipate its customers’ behavior and try to prevent the end of their cooperation, i.e., to predict the churn of customers.

Abstract: Digital educational tutorials are an inevitable part of the contemporary learning process, which is today typically organized in an online or blended environment. Creating and maintaining such tutorials classically, using video editors, is error-prone and time-consuming. We propose a more controllable approach for creating and reproducing educational tutorials by using Robotic Process Automation (RPA). Instead of recording videos, as in the classical approach, tutorials are created by specifying a sequence of user operations, which is executed using RPA. A teacher can later easily modify the tutorial just by editing this sequence. As well, since RPA can execute the operations directly on a student’s computer, they could get a more interactive tutorial than in the case of prerecorded video files. We have used the UiPath RPA tool for the creation and execution of the tutorials. As a case study, we created 8 educational tutorials for learning Java programming language.

Abstract: Industry 4.0 integrates technologies, organizational concepts and management principles in order to establish the flexible production, enabling mass customization of products. Smart factories, characteristic for Industry 4.0, produce large volume of various data. The data has to be handled and analyzed effectively and efficiently in order to adequately support decision making processes and production system's adaptability to frequent changes. Big data technologies are aimed to enable storage of large volumes of structured, semistructured and unstructured data alongside cost-effective, innovative forms of information processing for enhanced insight and decision making. The goal of this paper is to conduct a systematic literature review on the role of big data in Industry 4.0 with the focus on different data sources, technologies used for handling data and benefits of data analysis.

Abstract: Customer segmentation is the marketing practice of grouping customers according to certain characteristics. This paper presents a thorough exploration of customer segmentation using machine learning techniques, Logistic Regression, and Support Vector Machine (SVM), applied to data obtained from a mall customers database. By labeling the customer groups and analyzing their characteristics to gain deeper insights into their shopping behavior and preferences, the goal is to develop targeted marketing strategies and allocate resources efficiently to meet the specific needs of each customer segment. Applying statistical analyses and data visualization techniques, the study seeks to derive valuable insights from the data and identify discernible patterns and trends. Utilizing logistic regression yields a remarkable model accuracy of 98%. Subsequently, we employ another machine learning technique for data classification, namely the Support Vector Machine, which achieves an equally notable accuracy of 96%. Using these classification models, potential customers can be effectively converted into loyal ones and enhance the satisfaction of existing customers through tailored marketing strategies for each segment. The research offers insights into effective strategies for distinct customer groups. Applying these methods in a business setting can yield valuable information, forming a basis for informed decision-making and improving customer relationships through customer relationship management strategies.

Abstract: The field of Scanned document image analysis has many challenges that lie in the diverse nature of document images and variation in their data representation. The aforementioned results in the fact that the field of document analysis still has a lot of room for improvement, especially by using newly developed techniques in the field of deep learning. In this paper, we will present an implementation of a solution for analyzing data from scanned images of contracts written in the Serbian language using Faster RCNN networks.

Abstract: A scholarly knowledge graph systematically organizes academic information, including researchers, publications, and institutions, fostering comprehensive exploration of academic domains. Exposing these graphs seamlessly is pivotal for enhanced accessibility and consistency in scholarly communication. This paper is dealing with challenges associated with automated navigation of scholarly content, emphasizing the disparity between landing pages' designs for humans and machines. Employing Signposting, a standardized solution, the machine-readable data into landing pages within the University of Novi Sad's Research Information System were integrated. Landing pages of journals’ articles, conferences’ articles, monographs, monographs chapters, PhD dissertations, researchers, and institutions were enriched with signposting links. Links to authors, institutions (PhD), full text (pdf for PhD), license, and DOI were added.