737 papers found .

621. An approach for processing data from NASDAQ stock exchange database

ICIST 2020 Proceedings Part II, 256-259
Radojičić Dragana, Kredatus Simeon
Abstract: This research is based on a data set that replicates the entire NASDAQ stock exchange, which is the second-largest stock market (measured by market capitalization of shares traded). Since there is a huge number of stock market events in each trading day which caused the update of the shape of the Limit Order Book, the data produced by the stock market is huge and demanding for processing. In order to do research with a huge data set, we propose a framework for data processing.

622. Assessment of websites

ICIST 2020 Proceedings Part II, 260-265
Paroški Milan, Dragutinović Branislava
Abstract: In this paper, on the basis of the criteria from the Guidelines for Web site development, state administration bodies, territorial autonomy bodies and local self-government units, web sites were evaluated and recommendations for their improvement were given. A statistical analysis was made and further recommendations for the survey were outlined.

623. Lexicographical Index Decision Variable in Pulse-Doppler Radar Pulse Burst Waveform Optimization

ICIST 2020 Proceedings Part II, 266-270
Jevtić Miloš, Zogović Nikola, Graovac Stevica
Abstract: In pulse-Doppler radar pulse burst waveform optimization problem, the combination of pulse repetition intervals can be substituted with the lexicographical index (LI), thus simplifying the decision space without losing any degree of freedom. This approach relies on an algorithm that generates a combination from its LI. We benchmark such an algorithm, ACM Algorithm 515 (A515), on a modern computer. We estimate that A515 would negligibly slow down the evolutionary algorithm previously used as a solver, thus making a strong case for the LI decision variable approach within the subject problem. INTRODUCTION

624. Application of a Generalized Fuzzy Model in PKI Architecture Determination

ICIST 2020 Proceedings Part II, 271-276
Prodanović Radomir, Vulić Ivan, Bogićević Dusan
Abstract: It is difficult to decide what the best PKI architecture is to be applied due to unspecified determination parameters. The PKI architecture content and a comparative analysis of advantages and disadvantages are not sufficient for the appropriate determination of PKI as it is founded on subjective criteria. Subjectivity in interpretation of the PKI quality contributes to non-specificity of input data. The objective of the research is to obtain an efficient working framework for the application of the natural linguistic in determination of PKI architecture by using the fuzzy logic. In order to reduce the subjectivity in determining the appropriate PKI architecture, the authors propose the working framework based on selected parameters and fuzzy logic. For PKI architecture determination purposes, the authors use the value of a global limitation fulfillment for each architecture obtained by the Generalized Prioritized Fuzzy Constraint Satisfaction Problem.

625. Tooth detection with small panoramic radiograph images datasets and Faster RCNN model

ICIST 2021 Proceedings Part I, 1-4
Zdravković Milan, Pešić Zoran, Pešić Pavle
Abstract: This paper deals with tooth detection in realistic situations, characterized with relatively small number (in this case, 114 panoramic radiographs) of images for training object detection models, with differing qualities (contrasts, lightness) and sizes. The objective of the work is to define the pipeline and framework of decisions in x-ray images object detection problems. Faster RCNN architecture is used for detection; the approach includes transfer learning, augmentation and normalization (Contrast Limited Adaptive Histogram Equalization) of x-ray images. Resulting model performance is comparable with or exceeding the state of the art in the field, with mean Average Precision for the test set of mAP=0.96-0.97.

626. Image fusion for obstacle detection optimization

ICIST 2021 Proceedings Part I, 5-10
Ćirić Ivan, Pavlović Milan
Abstract: This paper presents a new approach for optimization of obstacle detection in railway infrastructure. The approach is oriented at the fusion of IR and night vision images, captured by the SMART obstacle detection system in real world scenarios. Those sensors can operate in low-light and at night conditions. However, due to different light and environmental conditions, object that is not visible on IR image may be visible on image from Night vision system. The fusion process includes artificial intelligence powered extraction and fusing data and features from IR image and image from Night vision system, as well as distance estimation between obstacle detection system and detected obstacles on railway infrastructure. Use of this approach can compensate hardware flaws of sensors and increase accuracy in obstacle detection in different conditions, and those increase safety and reliability of obstacle detection system.

627. Ontology-Driven Approach for Evidence Admissibility in Network Forensics

ICIST 2021 Proceedings Part I, 11-16
Matijević Milica, Gostojić Stevan
Abstract: Digital forensic investigators are commonly involved in legal proceedings aiming to support decision making. Therefore, the investigator’s testimony must rely on admissible digital evidence without grounds for impugning. This paper presents an ontology-driven application for assisting digital investigators to conduct a network investigation that leads to court admissible results. The application also enables collaboration among digital forensic experts, thus maintaining awareness of the latest technology, techniques, and tools.

628. Mobile Application for Quantitative Climbing Level Assessment

ICIST 2021 Proceedings Part I, 17-21
Volontar Matija, Kos Anton, Umek Anton
Abstract: Climbing in general, and especially sport climbing, has become an increasingly popular recreational and competitive activity. Consequently, there are many climbers who would benefit from quantitative insight into their climbing level. After reviewing the research area and existing solutions, we found there are many climbing-specific measurement systems that allow acquisition of quantitative data, from which climbing level can be assessed. The vast majority of measurement systems of this type are scientifically oriented and therefore cannot be used by the general public. The motivation for our work was the development of a multi-purpose and portable climbing-specific prototype measuring system for obtaining quantitative force data with the use of upper limbs and a design of a mobile application for managing the measurements and presenting their results. The measurement systems offers force measurements for a variety of climbing-specific grips. Evaluation of the measuring system was performed with twelve subjects of difierent climbing abilities. Within evaluation, we focussed on the user experience with the mobile application and on quantitative data obtained by the measuring system.

629. Online testing platform with eye tracking

ICIST 2021 Proceedings Part I, 22-26
Segedinac Milan, Savić Goran, Majić Igor
Abstract: In this paper we present a platform for online testing which tracks students’ gazes. EyeTrackQuiz enables a teacher to develop quizzes where every individual question can be fine-tuned via the editor which supports freely positioning of different elements (titles, introduction paragraphs, images, possible answers and code fragments). Furthermore, the editor supports defining regions of interests (ROIs). These regions represent rectangle regions of the screen, which the teacher considers important. While the student is taking the test, his eye movements is tracked by an eye tracker which is connected to the computer. Only gazes which fall into the regions of interest are included in the final reports. After the test conclusion, it is possible to see the results and eye tracking reports. Such reports give a better insight into student’s cognitive processes. The solution is implemented as an in-browser application, with an integrated software library for communication with the eye tracker.

630. Usability of Micro-Credentialing Functionalities

ICIST 2021 Proceedings Part I, 27-32
Dimitrijevic Sonja, Devedžić Vladan
Abstract: In the last few years, micro-credentials have gained in importance and popularity as small, stockable and shareable (on social media) achievement recognitions. The beginnings have not been easy though. Micro-credentialing technologies in their infancy have shown many limitations including those related to usability, which may have even slowed down their adoption. The objective of this paper is to present some unpublished results of a study on the usability of a couple of micro-credentialing platforms from five years ago. On this basis, other studies available in literature have been searched for, analyzed, summarized and critically evaluated in order to get an insight into how usability of the given technologies has been addressed over time. Some common usability pitfalls and potentially promising approaches on how to recognize and avoid such pitfalls have been identified, analyzed and discussed. Accordingly, the results of this paper are beneficial to current and new adopters of micro-credentials, whether they are issuers, learners/candidates or technology developers.

631. Empirical evaluation of Machine Learning models for Intrusion Detection

ICIST 2021 Proceedings Part I, 33-37
Takač Tomislav, Sladić Goran, Kovačević Aleksandar, Slivka Jelena
Abstract: Intrusion detection is vital in ensuring network security. Recently, researchers started focusing on developing machine learning (ML) techniques for intrusion detection as they promise to solve the problems of other techniques: high false positive alarm rates and the inability to detect unknown attacks. For the ML-based intrusion detection model to be useful in practice, it needs to be trained on the dataset that reflects a realistic network setting. Furthermore, as attacks evolve rapidly, it is crucial to use recent datasets that reflect current network environments. Recently, a publicly available dataset UNSW-NB15 was published. This dataset may serve as a recent realistic benchmark for training and evaluating ML models for intrusion detection. Most existing ML-based intrusion detection solutions have been evaluated on older benchmarks, and it is essential to evaluate how these approaches perform on UNSW-NB15. Thus, in this paper, we examine the literature to find which ML algorithms are frequently used for intrusion detection and evaluate them on UNSW-NB15. We discuss and compare our results to other papers using the same dataset.

632. Enabling visual analysis of tags hierarchy and usage within open data portals

ICIST 2021 Proceedings Part I, 38-43
Frtunić Gligorijević Milena, Bogdanović Miloš, Veljković Nataša, Stoimenov Leonid
Abstract: Data openness and transparency initiatives have led to a large amount of data being published on open data portals. These portals are focused on making the published data both accessible and discoverable. However, it is not a rare situation that metadata is incomplete, making it difficult for the users to obtain the desired information. To improve the discoverability, it is important to categorize datasets with missing category based on the available information. To do so, it is crucial to understand the usage and hierarchy of datasets’ metadata elements, particularly tags. In this paper we want to address this issue by introducing a tool for interactive visual analysis of the tags’ usage as well as the hierarchy of datasets metadata tags that describe different datasets in a single category of datasets. The tool we present relies on Formal Concept Analysis for creating a hierarchy of the usage of tags and combinations of tags used to describe different datasets. Furthermore, within the tool, we provide visualization of concept lattice that enables analysis of the links between the tags and determining the importance of a tag and a combination of tags within a category.

633. Multi-threading capability evaluation of the Notification Oriented Programming Language for the x86 Architecture

ICIST 2021 Proceedings Part I, 44-49
Kaehler Martini Guilherme Henrique, Ronszcka Adriano Francisco, Fabro João Alberto, Simão Jean Marcelo
Abstract: This short paper presents an initial evaluation of the Notification Oriented Programming Language (NOPL) to support multi-threading on x86 computers. This is taken as a necessary advancement to make such language of this new paradigm scalable and usable on modern platforms. The results show that it is possible to achieve such capability. Moreover, the initial numbers indicate that the research should continue. Further work is required to achieve commercial and industrial standards. Keywords multi-threading, computing efficiency, notification oriented paradigm, programming language.

634. Automatic Transformation of Plain-text Legislation into Machine-readable Format

ICIST 2021 Proceedings Part I, 50-55
Cvejić Andrija, Grujić Katarina-Glorija, Cvejić Aleksandar, Marković Marko, Gostojić Stevan
Abstract: Legislative documents are an important class of legal documents that regulate almost every area of people’s lives. The introduction of modern information technologies in the legal domain brings numerous benefits for the legal profession, but the lack of a machine-readable representation of legal documents brings some difficulties. This paper proposes a method for a regulation transformation from the plain text format into the machine-readable Akoma Ntoso format. The XML tagging process is implemented in three layers: metadata layer, structural layer and text layer. To apply structural and semantic markup, we use a combination of rulebased methods and methods based on neural networks. The output of our system is XML documents that comply with the Akoma Ntoso schema. These documents are technologyneutral and machine-readable.

635. Application of generic programming for the development of a C++ framework for the Notification Oriented Paradigm

ICIST 2021 Proceedings Part I, 56-61
dos Santos Neves Felipe, Simão Jean Marcelo, Ribeiro Linhares Robson
Abstract: Modern software development mainly relies on the use of the Imperative Paradigm (IP) and Declarative Paradigm (DP). However, despite widespread use of IP/DP, they have their drawbacks, such as the presence of structural and temporal redundancy, as well as cod coupling. As an alternative, the Notification Oriented Paradigm (NOP) introduces a new approach for developing software, based on the use of small reactive notifiable entities. Therefore, the NOP facilitates software development and allows achieving features such as responsiveness by redundancy avoidance and distributiveness by code decoupling. Some frameworks have been implemented to allow the development of NOP software applications in various programming languages, thereby changing the usual modus operandi of these languages. However, those NOP frameworks present some shortcomings, such as type limitation. This paper evaluates such shortcomings and proposes a new framework version based on generic programming, called NOP C++ Framework 4.0, showing how it improves in performance and usability over the most stable existing framework, the NOP C++ Framework 2.0.

636. An overview of Hyperledger blockchain technologies and their uses

ICIST 2021 Proceedings Part I, 62-65
Milićević Vladimir, Jović Jovana, Zdravković Nemanja
Abstract: Hyperledger is an open source community focused on developing a suite of stable frameworks, tools and libraries for enterprise-grade blockchain deployments. It presents an umbrella term comprising different blockchain technologies (BCTs), including distributed ledgers, libraries, and tools. Due to the variety of technologies, it is often not so easy to choose the correct BCT depending on the needs of a blockchain-based application. In this paper, we present an overview of the most (and less) popular Hyperledger distributed ledgers, comparing their similarities and differences between themselves and with Bitcoin and Ethereum, with the ultimate goal of simplifying the choice of ledger for a given application.

637. A generative model for the creation of a synthetic dataset for semantic segmentation

ICIST 2021 Proceedings Part I, 66-71
Vidović Mladen, Nešić Nebojša, Radosavljević Ivan, Mitrović Aleksandra, Obradović Đorđe
Abstract: The acquisition of large, annotated image datasets, required for the training of semantic segmentation models, is often an arduous task. This is because of the timeconsuming, complicated and error-prone nature of the process of manual image labelling. This process also often requires specialized software and domain knowledge. These problems can be circumvented by utilizing a generative model to create synthetic automatically labelled datasets. In this paper, we propose a generative model in the form of a 3D scene, representing an urban environment. A virtual camera setup is used to acquire labelled images from the virtual urban environment. Each image is stored as a multichannel EXR file, containing RGB data as well as an additional channel for each object class. These channels contain binary values which indicate whether a pixel belongs to the target class. These images are used to form a dataset for the training of semantic segmentation models. The viability of the generated dataset is evaluated by testing the trained semantic segmentation model on real world manually annotated images.

638. Experiment-driven system for machine learning based research management

ICIST 2021 Proceedings Part I, 72-76
Radosavljević Ivan, Obradović Đorđe, Konjović Zora, Mitrović Aleksandra, Gavrić Stanko, Vidović Mladen, Nešić Nebojša
Abstract: Due to the recent rapid expansion of machine learning and its more frequent use as a valuable tool in scientific research the lack of tools that incorporate research project management and tools for machine learning is becoming apparent. Although there are many tools that can be used for project management or machine learning experiments, most of them are intended for commercial use and usually do not support features necessary for scientific research such as experiment versioning, reviewing and publishing of the experiment results. In this paper the authors propose a software solution that provides a unified environment for running machine learning experiments and research project management. The solution is a microservice oriented web platform that can be used to manage research projects, catalogue datasets, define, validate, and execute machine learning experiments. Some of the advantages of the presented solution are its ability to integrate experiments directly into the research project management process, and to automatically version and validate those experiments with respect to the constraints imposed by the authors or reviewers of the experiments. An additional advantage of the solution is the use of the graphical DSL to define experiments, thus allowing researchers who are not skilled in programming to use the platform.

639. Providing End-user Oriented OBDA over Existing Relational Databases

ICIST 2021 Proceedings Part I, 77-81
Jeremić Aleksandar, Čeliković Milan, Dimitrieski Vladimir, Kordić Slavica, Luković Ivan
Abstract: Ontology-based data access (OBDA) is one possible approach to data management. In OBDA, the data is stored in external sources, such as the databases or files. The end-user view of the data is realized though ontologies, and queried as such. In the case of relational data sources, ontologies abstract away the technical details relating to the relational data model, and provide end-user with a more intuitive, object-oriented view of the data. The data layer of such systems may use many heterogenous data storage mechanisms. Relational databases are the most common, but other data storage mechanisms such as NoSQL databases, structured and semi-structured files may be used as well. The view layer of OBDA system is usually implemented though OWL ontologies, queried via SPARQL language. OWL ontologies in OBDA systems only serve to provide the conceptual view of the data stored in the database. Therefore the ontology only contains concept definitions, and not the data itself. This separation of the data (in the database) and its descriptions (in the ontology) imposes some challenges in the development of OBDA systems. For example, mappings need to be created that will connect data definitions in the data sources to the elements of the ontology. Also, using these mappings to accurately and efficiently answer end-user queries is an intricate task in itself. In this work, we present an automated system for bootstrapping a generic OBDA architecture, based on the ontop tool. Our system is able to generate the ontology that will serve as a conceptual view of the provided data sources, and map the data in the data sources to the elements of the generated ontology. An open-source tool ontop is used to subsequently provide a SPARQL endpoint for quering the data over the generated ontology. We put emphasis on generating an ontology that will describe the underlying data in a best possible way, by using database constraints to refine concept descriptions. Also, we provide a web-based front-end application, to aid in understanding and verifying the generated ontology by the end-user.

640. Conditionally Autonomous Drive from a Driver’s Perspective: A Survey

ICIST 2021 Proceedings Part I, 82-86
Gruden Timotej, Jakus Grega
Abstract: Since one of the key factors for successful introduction of conditionally autonomous vehicles is a properly designed user interface (UI) for take-over requests (TOR), we believe a systematical approach including drivers’ opinion and preferences should be pursued. We have conducted a short online survey with 126 participants asking for drivers’ preferences of UI modality, level of transferred information and vehicle behavior during TOR. The participants’ answers were in favor of auditory UIs (90%) and followed the “simple is the best” principle – preferring only a single alert, i.e., a beep (37%), for TOR.


Spread the word!



Eventiotic is a platform for provision of the unique digital services to scientific events' organizers, participants, and other interested communities.

Eventiotic repository facilitates universal and unrestricted access to the scientific events' online proceedings, for better visibility, more citations and more collaborations.