University of Niš: Facta Universitatis (E-Journals) / Универзитет у Нишу
Not a member yet
4905 research outputs found
Sort by
OPTIMISING PROCESS AUTOMATION OF GEOSPATIAL DATA PIPELINES BY ARTIFICIAL INTELLIGENCE
Full text linkThis study aims at using advanced GeoAI tools for monitoring landscapes in Italy using machine learning (ML) methods for remote sensing (RS) data processing. Changes in land cover types were identified using AI-processed satellite images. The methodology is based on the four ML algorithms of Python library Scikit-Learn embedded in the GRASS GIS: SupportVectorMachine (SVM), Decision Tree Classifier (DTC), RandomForest (RF) and Multilayer Perceptron Classifier (MLPC) of Artificial Neural Network (ANN). The multispectral satellite Landsat imagery was processed and analysed for changes in categories. The workflow of image processing includes classification for automatic detection of land categories. The presented maps demonstrated spatio-temporal vegetation dynamics and changes in land cover types detected using times series of the RS data. The topology of patches was detected by ML considering differences among spectral reflectance of pixels. ML algorithms recognised. Streamlined workflow through integration of RS and ML algorithms for model training, prediction and classification in GRASS GIS environment. This study has shown the advantages of AI methods for automation of RS data processing
A COMPARISON OF TOKENIZER PERFORMANCES – THE CASE OF SERBIAN LEGAL DOMAIN ADAPTATION
Full text linkThe advancement of large-scale language models in recent years has significantly enhanced various natural language processing (NLP) domains. This research addresses the specific challenge of developing BERT-based models tailored for domain-specific language modeling. Tokenization efficiency within BERT-based models is directly related to model efficiency which in turn motivated the research we present in this paper. Our goal was to enhance the development of a domain-adapted large language model for the case of Serbian language. Particular attention was given to the influence of training dataset content on the quality of the domain-adapted large language model tokenizer developed for masked language modelling. In this paper, we will present a comparison of the tokenization performance for a domain-adapted tokenizer in version 2 of SrBERTa language model we developed, against the performances of previous tokenizer version of SrBERTa model and two additional tokenizers of BERT-based models developed for Serbian language – BERTić and Jerteh-81. The results are generated using a test dataset consisting of 328482 samples of legal texts gathered from the Official Gazette of the Republic of Serbia
ANALIZA LEADER PRISTUPA U ODRŽIVOM LOKALNOM RAZVOJU U RUMUNIJI. STUDIJA SLUČAJA: GAL ZONA SĂTMARULUI
No full textThe LEADER approach is recognized as a pioneering methodology enabling rural communities to design and implement local development strategies. Originally a bottom-up initiative under the European Agricultural Fund for Rural Development, LEADER has evolved in Romania to address social inclusion, economic disparity and environmental protection. This article analyzes the LEADER approach within Romanian sustainable local development, evaluating the methodology's effectiveness by focusing on Local Action Groups (GAL) implementation and outcomes in the North-West Development Region through a case study of GAL Zona Sătmarului. By synthesizing multiple research sources, the study explores strategic planning, resource allocation, and performance outcomes of LEADER-initiated development strategies. The investigation examines how LEADER principles translate into practical solutions, assessing achievements and limitations in the Romanian context. This research contributes to broader sustainable rural development discourse, offering insights into challenges and opportunities facing Romanian communities pursuing locally driven development initiatives.Pristup LEADER se prepoznaje kao pionirska metodologija koja omogućava ruralnim zajednicama da osmisle i implementiraju lokelne strategije razvoja. Prvobitno inicijativa „odozdo nagore“ u okviru Evropskog poljoprivrednog fonda za ruralni razvoj, LIDER se razvio u Rumuniji kako bi se rešio problem socijalne inkluzije, ekonomske nejednakosti i zaštite životne sredine. Ovaj rad analizira LIDER pristup u okviru održivog lokalnog razvoja Rumunije, procenjujući efikasnost metodologije fokusirajući se na implementaciju i rezultate Lokalnih akcionih grupa (GAL) u Severozapadnom regionu razvoja kroz studiju slučaja GAL Zona Sătmarului. Sintetizujući višestruke istraživačke izvore, studija istražuje strateško planiranje, raspodelu resursa i rezultate razvojnih strategija pokrenutih pod pritiskom LIDER-a. Istraživanje ispituje kako se principi LIDER-a prevode u praktična rešenja, procenjujući dostignuća i ograničenja u rumunskom kontekstu. Ovo istraživanje doprinosi širem diskursu o održivom ruralnom razvoju, nudeći uvid u izazove i mogućnosti sa kojima se suočavaju rumunske zajednice koje sprovode lokalno vođene razvojne inicijative
UTICAJ ANGAŽOVANOSTI POTROŠAČA NA LOJALNOST BRENDU – POSREDNI EFEKAT ZADOVOLJSTVA ZAJEDNICOM
No full textIn the contemporary online environment, building consumer loyalty has become one of the key objectives of every company. An effective approach to achieving this goal is the development of an online brand community, which through the involvement of consumers enables their active communication and formation of brand-related attitudes. In this context, marketing managers recognise the importance of encouraging consumers to actively participate in the community, thereby further strengthening their relationship with the brand and increasing the likelihood of long-term loyalty. This study aims to examine the impact of consumer engagement in online brand communities on key marketing performance: community satisfaction and brand loyalty. Direct relationships were established between consumer engagement in the online brand community, community satisfaction and brand loyalty. The model is based on structural equation modeling (SEM), while the software program AMOS was used to measure SEM. The research results showed that consumer engagement in the online brand community has a statistically significant and positive impact on community satisfaction and brand loyalty. Additionally, it was found that community satisfaction is a predictor of brand loyalty.U savremenom onlajn okruženju, izgradnja lojalnosti potrošača postaje jedan od ključnih ciljeva svakog preduzeća. Efikasan pristup u ostvarivanju tog cilja jeste razvoj onlajn zajednice brenda, koja kroz uključivanje potrošača omogućava njihovu aktivnu komunikaciju i formiranje stavova prema brendu. U tom kontekstu, marketing menadžeri prepoznaju važnost podsticanja potrošača na aktivno učešće u zajednici, čime se dodatno učvršćuje njihov odnos sa brendom i povećava verovatnoća dugoročne lojalnosti. Ova studija ima za cilj da ispita uticaj angažovanosti potrošača u onlajn zajednicama brendova na ključne marketing performanse: zadovoljstvo potrošača zajednicom i lojalnost potrošača brendu. Uspostavljene su direktne veze između angažovanosti potrošača u onlajn zajednici brenda, zadovoljstva potrošača zajednicom i lojalnosti potrošača brendu. Model se zasniva na modeliranju strukturalnih jednačina (SEM), dok je softverski program AMOS korišćen za merenje SEM-a. Rezultati istraživanja su pokazali da angažovanost potrošača u onlajn zajednici brenda ima statistički značajan i pozitivan uticaj na zadovoljstvo potrošača zajednicom i njihovu lojalnost brendu. Dodatno je utvrđeno da je zadovoljstvo potrošača zejednicom prediktor lojalnosti potrošača brendu
OPTIMIZING CARDIOVASCULAR DISEASE DIAGNOSIS: A META-HEURISTIC AND FUZZY LOGIC-BASED APPROACH
Full text linkCardiovascular disease (CVD) is one of the most common and major global health challenges, which requires improved methods for early and precise detection and intervention. So, to recognize these heart problems and avoid sudden cardiac arrest, it is essential to detect abnormal heart conditions early. Machine learning (ML) based medical treatments are being implemented that are very helpful in quickly and effectively diagnosing CVD problems. One method that can offer practical answers to these kinds of problems is a meta-heuristic approach. Owing to its effectiveness, meta-heuristic approaches are presently used with medical data to diagnose conditions more practically and successfully than the traditional ML methods. In this study, we used three different meta-heuristic algorithms which are Genetic Algorithm (GA), Cuckoo Search Algorithm (CSA) and Particle Swarm Optimization (PSO) for diagnosis of the CVD diseases using two different datasets – CVD and Framingham. Finally, various ML classifiers were applied on the best selected features for both the datasets, obtained from the meta-heuristic algorithms for finding efficiency and comparing the results. The results demonstrate that Framingham dataset gives best accuracy of 98.47% by using CSA algorithm for feature selection and Random Forest as classifier whereas for the CVD dataset gives best accuracy of 94.12% by using PSO algorithm and Random Forest as classifier. Then, the best performing model is passed through some fuzzy logic rules to improve the model accuracy and gives better prediction for CVD prediction
HIGH-VOLTAGE SURGE IMPACT ON THICK-FILM SENSORS FOR STRUCTURAL HEALTH MONITORING: RESISTANCE AND NOISE SPECTROSCOPY ANALYSIS
Full text linkThis study explores the effects of high-voltage electrical surges on the performance and structural integrity of thick-film strain sensors developed for structural health monitoring in steel infrastructure. The sensors were fabricated using screen-printing techniques with a bismuth lead ruthenate-based resistive composition deposited on alumina ceramic substrates. To simulate realistic operational conditions, the sensors were mounted on steel beams and subjected to four-point bending to induce mechanical strain. Following mechanical loading, controlled high-voltage surge pulses were applied to emulate extreme electrical events. Sensor response was characterized before and after surge exposure using both static resistance measurements and current noise spectral analysis. While resistance measurements showed limited change, noise spectroscopy revealed microstructural damage undetectable by conventional means. The findings highlight the degradation mechanisms arising from electromechanical stress and demonstrate the effectiveness of noise spectroscopy as a non-destructive diagnostic tool. These results support the use of thick-film sensors in electrically demanding environments
OPTIMAL VAR PLANNING IN POWER SYSTEMS INCORPORATING DOMESTIC LOAD MODELING USING CHAOTIC TRIGONOMETRIC SEARCH ALGORITHM
Full text linkIn the last couple of years, with the advent of emerging load types, the global electricity markets have been witnessing fast-changing consumption behaviors. Across nearly all load sectors, modern nonlinear power electronic loads constitute a significant portion of the total electricity demand. It is becoming a challenging task for network engineers to maintain an uninterrupted power supply without compromising the network's efficiency and grid controllability. The proposed work presents an innovative planning strategy for optimal reactive power allocation in static load models. This load model utilizes exponent-based representations of active and reactive power, considering the seasonal and temporal variations to simulate their impact on transmission network flows. The objective is to minimize the overall operating cost while adhering to constraints imposed by the network. To ensure economic efficiency, the overall operating cost incorporates various components associated with VAr generation, along with the impact of transformer tap settings. The optimal parameters subjected to reactive power planning (RPP) are obtained by using the proposed CTSA (Chaotic Trigonometric Search Algorithm). The effectiveness of the proposed approach is validated on the Indian Utility 62-bus test system. Simulation results show a reduction in overall operating cost across all considered scenarios, demonstrating the efficacy of the proposed method in reactive power management and highlighting the superior performance and versatility of CTSA in addressing diverse operational challenges
PRIMENA VASPITNIH NALOGA U PRAKSI U REPUBLICI SRBIJI
No full textEducational orders are sui generis measures, which means that they are not criminal sanctions but instruments for redirecting social reaction towards juvenile offenders by attempting to transform social reaction into primarily non-penal responses. For the purposes of this paper, the data on the application of educational orders in the Republic of Serbia have been compiled from the annual reports of the Social Welfare Centers, as presented in the Annual Reports of the National Institute for Social Protection of the Republic of Serbia. On the basis of the collected data, the author draws conclusions on the trend of applying educational orders in Serbia. The subject matter of this research is an analysis of the application of educational orders towards juveniles on the territory of the Republic of Serbia in the period from 2018 to 2022, with the aim of determining the frequency of applying such educational orders.Vaspitni nalozi predstavljaju sui generis mere, odnosno nisu krivične sankcije već instrumenti preusmeravanja društvene reakcije prema maloletnim učiniocima krivičnih dela pokušajem njenog preobražaja u primarno nekazneno reagovanje. Podaci o primeni vaspitnih naloga na osnovu kojih je izveden zaključak o trendu primene vaspitnih naloga su izvedeni iz godišnjih izveštaja Centara za socijalni rad prezentovanim u Godišnjim izveštajima Republičkog Zavoda za socijalnu zaštitu za Srbiju (skr. RZSZ), Predmet istraživanja ovog rada predstavlja analiza primene vaspitnih naloga prema maloletnicima na teritoriji Republike Srbije u periodu od pet godina od 2018 – 2022. godine, sa ciljem da se utvrdi učestalost primene vaspitnih naloga
PRINCIPLES AND PHILOSOPHY OF OCCUPATIONAL HEALTH AND SAFETY MANAGEMENT – BASIC FEATURES
Full text linkThe need for occupational health and safety (OHS) is an ongoing one. OHS is generally defined as the science of predicting, identifying, assessing and controlling hazards and risks arising in or from the workplace that could adversely affect the health and well-being of workers and others at work. This takes into account the potential impact on the local community and the wider environment. OHS management is of great importance and is based on principles and philosophy. OHS principles relate to the rights, responsibilities and obligations of stakeholders who have certain expectations. This includes moral, social, legal and economic issues. The OHS philosophy begins with the basic need or requirement for safety, its foundations, analysis and assessment of causes, identification of unsafe conditions and actions and the sequence of events leading to an accident. It examines principles and methodologies for preventing accidents and implementing control measures promptly. The philosophy of OHS critically examines the basic principles and concepts of this field as an integral part of business management, while emphasizing their practical application to promote a safe and healthy work environment
EFEKTI SPRINTERSKOG TRENINGA SA OPTEREĆENJEM I TRADICIONALNOG SPRINTERSKOG TRENINGA NA TENZIOMIOGRAFSKE PARAMETRE MLADIH FUDBALERA
No full textThe main goal of this study was to examine the effects of an 11-week weighted vest sprint training (WVT) program and traditional sprint training (TST) on tensiomyography (TMG) parameters in young football players. Fifty young football players were divided into two groups: a group that performed WVT and a group that stopped the running session if the obtained time was higher than 3% (WVG - n=25; age: 17.92±1 yr; body mass: 71.25±7.9 kg; height: 178.79±6.4 cm) and a group that performed TST (TSG - n=25; age: 17.92±1; mass: 75.49±5.56; height: 178.75±3.83). The experimental program lasted for 11 weeks and was implemented twice a week at the start of the training sessions separated by at least 48 h. Anthropometry (standing height, sitting height and leg length) was measured, as well as body composition, on the basis of which graduation was later calculated, which served as a covariate. After that, the TMG parameters of m. rectus femoris, m. vastus lateralis, m. vastus medialis and m. tibialis anterior were assessed. Time × group effects were examined with the Bayesian linear mixed-effects ANCOVA model. Between‐group difference was expressed as the posterior mean standardized regression coefficient (β) with its 95% credible interval (95% CI) for each variable. Practical significance was evaluated using the Region of Practical Equivalence (ROPE). The results showed that the applied programs did not significantly contribute to changes in the TMG parameters, which indicates that this type of training stimulus may not have a sufficiently specific effect on the neuromuscular characteristics that can be measured by TMG.Glavni cilj ovog istraživanja bio je da se ispitaju efekti sprinterskog treninga sa opterećenjem i tradicionalnog sprinterskog treninga na parametre tenziomiografije (TMG) mladih fudbalera. Pedeset mladih fudbalera je podeljeno u dve grupe: grupu koja je imala sprinterski trening sa prslucima sa tegovima i prekidala serije kada je mereno vreme sprinta bilo veće za minimum 3% od najboljeg vremena na tom treningu za tu seriju (WVG – n=25; starost: 17.93±1 god.; telesna masa: 71.25±7.9 kg; telesna visina: 178.79±6.4 cm) i grupu koja je imala tradicionalni sprtinerski trening (TSG – n=25; starost: 17.92±1 god.; masa: 75.49±5.56 kg; telesna visina: 178.75±3.83 cm). Eksperimentalni program je trajao 11 nedelja i sprovođen je dva puta nedeljno na početku treninga razdvojenih najmanje 48 h. Merena je antropometrija (stajaća visina, sedeća visina i dužina nogu), kao i telesni sastav, na osnovu kog je kasnije izračunata maturacija koja je služila kao kovarijata. Nakon toga, procenjeni su parametri TMG mišića m. rectus femoris, m. vastus lateralis, m. vastus medialis i m. tibialis anterior. Efekti vreme × grupa ispitivani su Bajesianskim linearnim ANCOVA modelom mešovitih efekata. Razlika između grupa izražena je kao posteriorni srednji koeficijent standardizovane regresije (β) sa 95 % intervalom poverenja za svaku varijablu. Praktični značaj je procenjen korišćenjem regioan praktične ekvivalencije (ROPE). Rezultati su pokazali da primenjeni programi nisu značajno doprineli promenama u parametrima TMG ni jednog mišića, što ukazuje da ova vrsta trenažnog stimulusa možda nema dovoljno specifičan efekat na neuromišićne karakteristike koje su merljive TMG-om