Abstract: Forest fires are causing great problems in Brazil where majority of the land is under the forests, with Amazon being the largest and most famous one, often referred to as the lungs of the world. A big step in solving the problem and thus reducing the damage caused by the forest fires would be if competent authorities could have an insight in which of the regions are the most threatened ones. This paper offers way to classify federal states of Brazil in different times of the year according to the level of vulnerability using several different methods for classification. Data set used data regarding socio-economical and temporal data and the idea was to see their influence on forest fires in federal states of Brazil during different periods of the year. Data included a lot of features, but the ones that attracted attention were the month and year in which the fire has occurred and the area of the federal state. Best results were obtained by using random forest algorithm for the classification problem. Some other algorithms like SVM were expected to give better results than they did, since they apply well on this kind of classification problems.

Abstract: This research delves into assessing the impact of hydromorphological parameters on macroinvertebrate communities within the Danube River basin, utilizing data from the Joint Danube Survey 2 expedition. By focusing on key factors such as substrate type, hydrological alterations, and macrophyte vegetation structure, an Artificial Neural Network (ANN) – based model was developed to accurately predict the ecological state, offering a promising tool for future river biodiversity monitoring aligned with hydromorphological characteristics. The focus of this modeling is on utilizing ANNs to model mollusc fauna in the Danube River, considering hydromorphological parameters. Addressing a gap in the literature, this research provides a specialized analysis, aiming to expand understanding of the complex relationship between mollusc taxa, and hydromorphological conditions in river basins. The ANN model utilized input variables including substrate type, flow rate, hydrological alterations, macrophyte vegetation structure, and bank modification types, and 42 taxa of mollusc fauna represented on a binary scale, as outputs. The evaluation of performance metrics, including precision (0.69), recall (0.52), F1 score (0.59), and accuracy (0.85), emphasizes the model's efficacy in predicting macroinvertebrate community structures based on hydromorphological parameters, offering valuable insights for ecological and environmental engineering endeavors.

Abstract: The research aims to assess the feasibility of estimating a broad spectrum of soil properties using multispectral satellite imagery and deep learning. Soil properties play a crucial role in agriculture, environmental conservation, construction, and other sectors. The study covers the entire USA in 2023, collecting data from geolocations with soil properties and monthly satellite images. Employing a convolutional model with ReLU activation and Mean Absolute Error (MAE) loss, the training process included pixel shuffling to enhance data variability. The results show good predictions for some soil properties, with a focus on Normalized Mean Error (NME), Normalized Mean Absolute Error (NMAE), and R² metrics. Combining images from different months yields improved results, and scaling predictions enhances accuracy. The study demonstrates promising outcomes, highlighting the potential for satellite imagery and deep learning in accurately estimating soil properties.

Abstract: Our work centers on the development of a robust Serbian Cyrillic Tokenizer, a crucial component for processing Serbian text from diverse sources. This tokenizer adeptly handles language-specific challenges such as morphological complexity and script variations, playing a pivotal role in democratizing access to advanced language models. The methodology encompasses thorough stages including data acquisition, detailed preprocessing, and comprehensive training. This holistic approach ensures exceptional data quality, which aligns with the rigorous specifications required for training advanced models effectively. Our tokenizer not only demonstrates enhanced performance in managing the morphological complexities of Serbian but also contributes significantly to the field of language-specific NLP tools. By improving the precision and efficiency of language processing, our project supports the development of more accurate and responsive language technologies, fostering innovation across various applications of NLP in academic and commercial sectors.

Abstract: This paper aims to describe the process of the implementation of a web application for a song recommendation engine based on lyrics analysis. The paper gives an overview of existing algorithms used in major streaming platforms, describing data and processes used for song recommendation generation, as well as exploring their imperfections. It continues to describe the methods used in the implementation of the solution, focusing on data extraction and processing, together with the aspects of web application development. Finally, the paper presents the result in a song recommendation engine which uses signals extracted from the song lyrics, implemented in a scalable and sustainable way.

Abstract: In recent years, we have witnessed the attempts of offering cloud services closer to data and end users. Even though there already are platforms on the market managing infrastructure at the edge, there are still many challenges yet to be resolved. One example would be serving requests from an optimal location. The upcoming real-time applications will be bound to rely on local data processing because of its sheer volume and strict time constraints, but this doesn’t completely exclude the cloud from the picture. It will still be the best suited option for resource-intensive workloads. In this paper, we present the distributed cloud model that enables dynamic formation of ad-hoc clouds. By abstracting infrastructure to the level of software, we manage to streamline the process of resource reallocation, so that even the highly fluctuating needs of nearby population can be met. We also discuss the idea of data transfer from one distributed cloud to another when user mobility is detected. This strategy aims to minimize distance of users from data at all times, but the particular difficulty we focus on is how such a process can be executed with respect to data privacy requirements.

Abstract: In this paper, a detailed literature review on the security issues and vulnerabilities of smart contracts used in blockchain systems is presented. Smart contracts automate and enforce agreements in a trustless environment such as public blockchain, security challenges become critical before blockchain deployment. Predominantly written in Solidity, a programming language designed for Ethereum, these contracts are pivotal in various applications, ranging from financial services to supply chain management. Our review starts with some unique characteristics of Solidity as a programming language. Afterwards, we present well-known and less-known security vulnerabilities such as reentrancy attacks, integer overflows, and underflows, and improper access control, which could lead to questionable reliability of decentralized applications running on the blockchain. The paper systematically categorizes these vulnerabilities, illustrated in real-world examples to highlight the potential risks and consequences, and examines the underlying causes of identified security flaws. Our results emphasize that certain vulnerabilities, ranging from language-specific pitfalls in Solidity to broader issues in smart contract design and deployment can lead to strengthening of automated security audits, formal verification methods, and best coding practices in enhancing the security posture of smart contracts.

Abstract: This article explores the advancement of artificial intelligence in bariatric surgery, envisioning a framework for understanding and synergy in the domain in a systemic way. We analyze the evolution of the topic from 1988 to 2024, using content analysis and meta-analysis methods, establishing a chronological perspective and emerging trends. The methodology adopted a systematic literature review based on the PRISMA protocol, collecting publications in the PubMed, Scopus and Web of Science databases from 1988 to 2024. Advanced analytical tools, such as the Biblioshiny Tool, were used to visualize and analyze progress. The results clarified the benefits and challenges of artificial intelligence in bariatric surgery, identifying key elements and analyzing health topics, offering future perspectives of machine learning, producing predictive models, deep learning, developing algorithms for identifying operative steps in various types of bariatric surgery, artificial neural networks, having more accuracy than machine learning, data mining, used in the prediction of information in weight loss. Limitations: The analysis was limited to the PubMed, Scopus and Web of Science databases, focusing on the application of artificial intelligence in bariatric surgery, without considering other health fields such as nursing, nutrition, speech therapy, physical education, etc. Originality: This study is a pioneer in SLR on the application of AI in bariatric surgery, integrating multidisciplinary expertise and advances in AI such as machine learning and deep learning, contributing to the field of digital health and bariatric surgery.

Abstract: The integration of technology into education has significantly increased after the COVID-19 pandemic. Pre-trained Large Language Models (LLMs), such as Generative Pre-training Transformers (GPT) [5] and Bidirectional Representations from Transformers (BERT) [4], with their capabilities for comprehending and generating natural language, have gained prominence in the area of modern Natural Language Processing (NLP) problems. In this paper, we present a knowledge assessment system based on natural language processing for the Serbian language. The system encompasses functionalities for the creation, storage and automatic evaluation of student’s tests, providing valuable insights for teachers. By utilizing LLMs, we determined the correctness of student’s free-form answers by comparing them with the correct teacher's answers. In assessing the semantic similarity of answers, we evaluated two approaches: LLMs embeddings with cosine similarity and ChatGPT, with ChatGPT producing the best results.

Abstract: This paper provides a comprehensive exploration of the application and significance of Seccomp profiles within distributed cloud environments, with a particular emphasis on investigating Seccomp profile trees. The primary focus is on analyzing the hierarchical organization of namespaces and the creation of Seccomp profile trees within a distributed cloud computing platform. Furthermore, the research includes an examination of Seccomp profile inheritance, along with the integration of access modifiers directly within security profiles. Additionally, the paper presents a proof-of-concept implementation to validate the proposed concepts.

Abstract: In this study, we developed a comprehensive approach to identify biomarkers for chronic inflammation (CI), combining machine learning (ML) with traditional bioinformatics. Differential expression analysis on dataset of 256 samples from CI model containing 54,675 genes, reduced the dataset to 1,623 genes based on adjusted p-value and log fold change criteria. These genes were then subjected to a Random Forest (RF) algorithm, achieving around 77% accuracy in distinguishing between control and inflammation groups. Concurrently, a gene enrichment analysis identified 184 genes significantly associated with inflammatory pathways, with the RF model showing similar accuracy. This dual-path approach was essential due to the complexity of the CI samples, which excluded cancer but included various CI diseases. The integration of these two analytical paths led to the identification of 36 key genes, meeting both differential expression and pathway enrichment criteria. This methodology not only advances our understanding of CI's genetic basis but also paves the way for future targeted research and therapeutic interventions in chronic inflammatory diseases.

Abstract: Drilling, a frequently employed machining procedure, utilizes a rotating cutting tool known as a drill to remove material from the workpiece. Wide applicability of drilling as a machining process, arose a lot of need for process and parameter optimizations. Significant challenges within the realm of drilling operations revolve around optimizing process parameters and devising tool routes. Various drilling route planning issues arise, encompassing single tool drilling, multi-tool drilling, multi-tool drilling with precedence constraints, and multi-tool hole drilling with sequence-dependent drilling times. In present research, one multi tool drilling path problem was decomposed to three single tool drilling path problems. Each of these problems were first solved with CAM software, then formulated as traveling salesman problem (TSP) and solved with metaheuristics method i.e., simulated aneling (SA). A comparison of achieved results was then conducted to find the shortest tool path for drilling machining operation. Tool path optimization with SA, achieved smaller drilling paths compared to drilling paths derived with CAM software. This optimization can generate improved hole drilling sequences, leading to significant savings in the time and overall distance traveled by the drill. This suggests the practicality of using it.

Abstract: The immense shift to cloud computing has brought changes in security and privacy requirements, impacting critical Identity Management (IdM) services. Currently, many IdM systems and solutions are accessible as cloud services, delivering identity services for applications in closed domains and the public cloud. This research paper centers on identity management in distributed environments, emphasizing the importance of robust up to date auhtorization mechanisms. The paper concentrates on implementing robust security paradigms to minimize communication overhead among services while preserving privacy and access control. The key contribution focuses on solving the problem of restricted access to resources in cases when the authentication token is still valid, but permissions are updated. The proposed solution incorporates an Identity and Access Management (IAM) server as a component that authenticates all external requests. The IAM server’s key responsibilities include maintaining user data, assigning privileges within the system, and authorization. Furthermore, it empowers users by offering an Application Programming Interface (API) for managing users and their rights within the same organization, providing finer granularity in authorization. The IAM server has been integrated with a configuration dissemination tool designed as a distributed cloud infrastructure to evaluate the solution.

Abstract: Exoskeleton-assisted passive rehabilitation provides a personalized, intensive and task-specific training with a reduced need for therapist supervision. This method leads to significantly improved functionality and reduced spasticity for post-stroke patients compared to traditional rehabilitation. However, the approach for position control of the exoskeleton in such a case often relies on simplified decoupled models that controls each joint separately to simplify the control problem. Consequently, even robust control strategies like Active Disturbance Rejection Control (ADRC) face limitations due to significant time-varying and user-specific parameter variations, leading to reduced tracking accuracy. To address these challenges, while maintaining simplicity in the control design process, an adaptive control strategy based on ADRC was proposed that dynamically estimates system parameters with Kalman filter. This strategy actively estimates system parameters online using Kalman filter, enhancing the robustness capabilities of the ADRC by providing up-to-date estimated model parameters. The effectiveness of the proposed method was demonstrated through simulated experiments on a 5-degree-of-freedom (DOF) upper limb exoskeleton model.

Abstract: In this paper, we present a lightweight distributed private blockchainbased framework for credentialing documents issued by higher education institutions (HEIs), such as diplomas and diploma supplements. By utilizing Hyperledger Fabric, the most popular distributed ledger technology for private blockchains, we propose a secure credentialing model comprised of three layers – the smart contract layer, the blockchain layer itself, and the private node and network layer. With a minimal number of functionalities such as issuance and verification, our lightweight system can be deployed on a trustful environment, e. g. faculties from the same university, or a consortium of universities. With such an environment, we eliminate the need for a computationally complex consensus mechanism for adding blocks to the ledger, while retaining easy implementation with the HEIs information system and/or learning management system. Based on previous research and prototyping, our model presents the next step in implementing a private blockchain for HEIs to add security to their issued documents, which can be easily validated.

Abstract: In recent time, social media has become a strong tool in the Nigerian political landscape, fostering interaction on governance and promoting active involvement of citizens in policymaking. The most disruptive policy in Nigeria since the Covid 19 era is the decision of the government to remove fuel subsidy in May 2023. Like many other policies of the government, this has been brought under public scrutiny and it has generated a lot of reactions on social media, particularly “X” Many applaud the decision while many argue it is an ill-conceived policy that will bring hardship to the masses. In this study, a substantial amount of data on Nigerians’ opinion on fuel subsidy removal is collected from X as a univariate dataset. In the first phase of the study, we classify the data using VADER and got an accuracy of 82.22% after evaluating the performance. In the second phase, we utilize the dataset to train a BERT model for binary classification, with the goal of enabling our model to classify future opinions on the subject as either positive or negative sentiment. With an average accuracy of 84.98% across different variants of BERT, we demonstrate the effectiveness of transformer-based model in identifying sentiments expressed by Nigerians regarding the fuel subsidy removal policy. Thus, the study contributes to Nigerian natural language processing studies.

Abstract: Digital transformation has generated profound structural changes in companies, impacting their products, processes, and services. It involves the integration of emergent technologies into all aspects of the companies, changing how work is done. From automating repetitive tasks to implementing data-driven decision-making processes, digital transformation optimizes workflows, fosters innovation, and drives competitive advantage. The study hypothesizes that effective technology implementation with employee engagement, usability awareness, and strategic management practices enhances digital transformation outcomes. Therefore, this research investigates the influences and correlations within four key domains: digital transformation, human factors, business process management, and emerging technologies. In this way, a literature review in the Scopus and Web of Science databases identified relevant articles aligning with the investigation's scope. The content analysis allowed us to determine elements and parameters that would serve to employ Structural Equation Modeling (SEM), a robust multivariate data analysis approach designed for exploring complex relationships. The results indicate that digital tools optimize operations and decision-making processes, aligning with user experience principles to mitigate cognitive overload. In addition, this research underscores the importance of considering human factors in technology implementation, aiming to guide companies in balancing efficiency with employee cognitive load. Finally, future work will propose the conceptual framework to ensure its practical applicability in enhancing human experience in digitally transformed organizations.

Abstract: Digital technologies have occupied a very important place in people's daily life and work for years. Their application is gaining more importance every day, and those technologies are very influential. It is safe to say that the application of digital technologies has introduced significant improvements in jobs that were performed in the traditional way. When it comes to the food processing process, there is an increasing number of research and examples coming from the industry that digitalization has introduced key innovations and principles into the process of food processing and preparation. The goal of this paper is to review the current state of digital technology use in the process of food processing and meal preparation. Some of the basic concepts covered are the application of sensory technology, augmented reality, artificial intelligence, mobile apps, machine learning, and etc. The second goal of this paper is to compare the advantages and disadvantages of using different digital technologies in gastronomy.

Abstract: The escalating adoption of patient-centric care paradigms and software-driven technologies within healthcare institutions signifies a departure from conventional methodologies. Current strategies for acquiring patient information prioritize understanding individuals' perspectives and experiences with their health conditions. In response, production engineering has embarked on rigorous research and methodological development, aiming to leverage this data for transformative innovations in healthcare provision. This study introduces a preliminary framework facilitating the seamless integration of health management and process mining methodologies, with a primary emphasis on elevating healthcare quality. Conducting a meticulous systematic literature review and content analysis, this research scrutinized 608 scientific works identified through strategic keyword combinations. The ensuing analysis categorized 148 articles based on predefined inclusion and exclusion criteria. Following the systematic review, 73 articles pertinent to the research topic were identified, with 18 articles meeting the established content analysis criteria. This comprehensive process not only revealed pivotal references but also identified existing gaps in the field, providing an opportunity to delve into patient-centered care and assess the level of care provision. The preliminary framework developed herein is designed to offer a systematic approach for evaluating health systems, with a specific focus on identifying areas for enhancement in healthcare service delivery. Serving as a foundational step, this preliminary framework lays the groundwork for the critical evaluation of health systems and the identification of aspects necessitating improvement throughout the provision of healthcare services.

Abstract: Being an artificial structure, a highway is bound to have an impact on the environment, especially when it is in proximity of watercourses and water bodies. Hydraulic engineers are responsible for proving that highway impacts will be avoided or mitigated to the greatest extent. The hydraulic analysis of the river and its interaction with numerous structures existent or planned in the river valley is performed to provide insights into possible impacts on the environment. To facilitate improved design, improved communication, and more effective design delivery, a collaborative hydraulics platform is established. Software for two-dimensional (2D) hydraulic modelling, graphical user interfaces, and auxiliary materials are examples of next-generation technologies. Today's engineers and designers may easily create and modify 2D models thanks to advancements in computer hardware, modelling software, Geographic Information Systems (GIS), and survey techniques. High visualization capabilities enable elaborate graphical output that helps communicate design outcomes and consequences to different stakeholders. A computational platform is essential for the execution of collaborative hydraulics since it links disparate software programs and facilitates efficient data interchange and communication amongst the specialists in a design team. The key components of the collaborative hydraulics technique are presented in this paper, along with some findings from its use on the Morava Corridor Project. Hydraulic models are being updated in accordance with the ongoing work on the Morava Corridor Project to be utilized going forward for the project's implementation, through design, construction, and maintenance.