This study applies the PIPRECIA-S model to evaluate and prioritize artificial intelligence (AI) applications in higher education. The aim is to identify which AI-driven solutions offer the greatest potential for improving teaching, learning and institutional management. Seven relevant criteria were defined based on the literature and expert consultation: pedagogical effectiveness, student engagement, adaptability, ease of implementation, data security, cost efficiency and long-term sustainability. Four AI applications were selected for evaluation: adaptive learning systems, intelligent tutoring systems, predictive analytics for student success and automated administrative tools. A panel of five domain experts independently provided pairwise comparisons to weight the criteria, while a separate panel assessed the alternatives. The results indicate that Adaptive Learning Systems achieved the highest overall score (4.375), followed by Intelligent Tutoring Systems (4.107) and Predictive Analytics (4.062), while Administrative Tools (3.647) ranked lowest. These findings highlight the pedagogical priority of AI solutions that directly enhance learning outcomes and student retention. The paper contributes by demonstrating how multi-criteria decision-making (MCDM) models, such as PIPRECIA-S, can support evidence-based policy and strategic planning in higher education, providing both theoretical insights and practical guidelines for sustainable AI integration.

Deep learning models, known as convolutional neural networks (CNNs), have paved the way for reliable automated image recognition. These models are increasingly being applied in research on freshwater biodiversity, aiming to enhance efficiency and taxonomic resolution in biomonitoring. However, insufficient or imbalanced datasets remain a significant bottleneck for creating high-precision classifiers. The highly imbalanced data, where some species are rare and others are common, are typical of the composition of most benthic communities. In this study, a series of CNN models was built using 33 species of aquatic insects, with datasets ranging from 10 to 80 individuals, to determine the optimal number of individuals each class should have to build a high-precision classifier. We also consider the effect of class imbalance in the training dataset and the use of oversampling technique. The results showed that a robust model with acceptable accuracy (99.45%) was achieved with at least 30 individuals per class. A strongly imbalanced dataset caused an approximately 2% decrease in classification accuracy, while a moderately imbalanced dataset had no significant effect. The application of the oversampling technique enhanced in 1.88% the accuracy of strongly imbalanced models. These findings can help effectively tailor future aquatic macroinvertebrate training datasets.

The identification of aquatic macroinvertebrates, particularly dark taxa like Chironomidae, due to their complex morphological features and unresolved taxonomy hinder the efficiency of routine biomonitoring. This study proposes an unsupervised deep clustering approach using β-variational autoencoders (β-VAEs) to identify chironomid larvae morphotypes in a completely unsupervised manner. A dataset of 5365 chironomid specimens from 37 taxa was used to develop and test multiple β-VAE models. The number of latent features (20–80) and the β hyperparameter (0.1–10) were systematically varied to optimize unsupervised classification accuracy. Loss analysis revealed that models with fewer latent features exhibited better feature disentanglement and reduced total correlation (TC) loss, enhancing the unsupervised classification of chironomid taxa. The model with 30 latent features and β = 0.1 outperformed others, achieving the highest Normalized Mutual Information (NMI) scores for clustering with K-means (0.4438) and Louvain (0.4813) algorithms. Entropy analysis revealed that species such as Diamesa insignipes, Rheocricotopus fuscipes, and Tvetenia tshernovskii posed classification challenges for the β-VAE model, as specimens from the same species were often assigned to multiple clusters. β-VAE showed in the present study the potential of unsupervised clustering for taxonomic identification, offering a scalable approach for biomonitoring programs. By enabling the identification in unsupervised manner, this study contributes to the inclusion of dark taxa in bioassessment and the exploration of cryptic diversity, advancing biomonitoring and biodiversity conservation.

The growing demand for deployment of Artificial Intelligence (AI) on resource-constrained edge devices has motivated extensive research on the design of efficient edge-compatible AI hardware accelerators. One of the most promising solutions are the self-adaptive AI accelerators, capable of optimizing in real time their performance and energy consumption according to application requirements. This work introduces the EU-funded project Twinning for Excellence in Adaptive Edge Artificial Intelligence (AIDA4Edge), aimed to advance the state-of-the-art in the design of adaptive neural network accelerators for edge applications. The main goal is to develop a novel hybrid self-adaptive neural network architecture combining spiking and artificial neural networks, and supporting runtime adaptation of network functionality, precision and reliability. Furthermore, we aim to enhance the neural network training by incorporating hardware and quantization constraints in an automated tuning engine.

Due to the increasing prevalence of irregular sampling across various scientific disciplines, the analysis and forecasting of irregular time series has gained growing importance in contemporary research. Unlike regular time series, irregularly sampled data introduces specific challenges, including temporal misalignment, missing values, variable-length observations, and data sparsity. Recent literature presents a wide range of methodologies developed to address these challenges, with particular emphasis on the use of Explainable Artificial Intelligence (XAI). The aim of this paper is to identify and analyze XAI models that explicitly address irregularities in time series, including those capable of combining interpretability with the capacity to learn from sparse, temporally inconsistent, and nonlinear data. Furthermore, we discuss the importance of optimizing these models for deployment on edge devices in line with Industry 5.0 principles, emphasizing human-centric and transparent decision-making processes.

Edge artificial intelligence (Edge AI) is increasingly vital in smart surveillance, industrial IoT, and autonomous systems, where real-time processing and energy efficiency are critical. Traditional deep neural networks are accurate but static and resource-hungry, limiting deployment on constrained devices. Dynamic neural networks enable input-adaptive computation, adjusting workload at runtime to context, input difficulty, and hardware limits. This survey synthesizes methods that support dynamic adaptability alongside offline compression (e.g., pruning, quantization) and runtime strategies (e.g., early exiting, dynamic resolution control). We review supporting software frameworks, including Slimmable Networks, Once-for-All (OFA), and PyTorch-based modules for adaptive execution. Finally, we outline a research agenda to co-optimize pre-deployment compression with runtime adaptation, targeting accuracy, latency, energy trade offs for robust, real-world edge deployment.

Artificial intelligence (AI) is becoming increasingly important in higher education, which has resulted in the accelerated development of research in this area. This paper conducts a bibliometric analysis of scientific papers researching AI applications in higher education, using the Web of Science database. The analysis covers the period from 1996 to February 2024 and focuses on the most cited works in this field, a total of 82 papers, with 1011 citations (944 without self-citations). Our analysis shows that interest in AI has increased significantly over the past few years, with the most dominant research in the fields of education, computer science, and engineering. The largest number of papers was published in 2023, which indicates the growing importance of this topic. These results provide a foundation for future research on the impact of AI on educational practices, its challenges, and its potential to transform education in the future.

This paper presents a comprehensive model for cyber security risk assessment using the PIPRECIA-S method within decision theory, which enables organizations to systematically identify, assess and prioritize key cyber threats. The study focuses on the evaluation of malware, ransomware, phishing and DDoS attacks, using criteria such as severity of impact, financial losses, ease of detection and prevention, impact on reputation and system recovery. This approach facilitates decision making, as it enables the flexible adaptation of the risk assessment to the specific needs of an organization. The PIPRECIA-S model has proven to be useful for identifying the most critical threats, with a special emphasis on ransomware and DDoS attacks, which represent the most significant risks to businesses. This model provides a framework for making informed and strategic decisions to reduce risk and strengthen cyber security, which are critical in a digital environment where threats become more and more sophisticated.

Bioassessment is the process of using living organisms to assess the ecological health of a particular ecosystem. It typically relies on identifying specific organisms that are sensitive to changes in environmental conditions. Benthic macroinvertebrates are widely used for examining the ecological status of freshwaters. However, a time-consuming process of species identification that requires high expertise represents one of the key obstacles to more precise bioassessment of aquatic ecosystems. Partial automation of this process using deep learning-based image classification is the goal of an ongoing project AIAQUAMI we are participating in. One of the project goals is to develop software support for image classification with visualization and reporting. For that purpose, we developed desktop and web applications that we open-sourced as Imagelytics Suite. Both desktop and web applications rely on a convolutional neural network (CNN) to classify images and the Grad-CAM algorithm to produce heatmaps of the image areas that mostly influenced the network decision. Along with the source code of the applications, we also open-sourced scripts that can be used to train CNN on an arbitrary dataset and produce required metadata, so it can be used with Imagelytics applications. In this article, we presented technical details regarding the design of the applications and the training method that will enable their general use for image classification tasks. As a part of the evaluation, we will show a use case related to species identification of non-biting midges (Diptera: Chironomidae).

Deep learning techniques have recently found application in biodiversity research. Mayflies (Ephemeroptera), stoneflies (Plecoptera) and caddisflies (Trichoptera), often abbreviated as EPT, are frequently used for freshwater biomonitoring due to their large numbers and sensitivity to environmental changes. However, the morphological identification of EPT species is a challenging but fundamental task. Morphological identification of these freshwater insects is therefore not only extremely time-consuming and costly, but also often leads to misjudgments or generates datasets with low taxonomic resolution. Here, we investigated the application of deep learning to increase the efficiency and taxonomic resolution of biomonitoring programs. Our database contains 90 EPT taxa (genus or species level), with the number of images per category ranging from 21 to 300 (16,650 in total). Upon completion of training, a CNN (Convolutional Neural Network) model was created, capable of automatically classifying these taxa into their appropriate taxonomic categories with an accuracy of 98.7 %. Our model achieved a perfect classification rate of 100 % for 68 of the taxa in our dataset. We achieved noteworthy classification accuracy with morphologically closely related taxa within the training data (e.g., species of the genus Baetis , Hydropsyche , Perla ). Gradient -weighted Class Activation Mapping (Grad -CAM) visualized the morphological features responsible for the classification of the treated species in the CNN models. Within Ephemeroptera, the head was the most important feature, while the thorax and abdomen were equally important for the classification of Plecoptera taxa. For the order Trichoptera, the head and thorax were almost equally important. Our database is recognized as the most extensive aquatic insect database, notably distinguished by its wealth of included categories (taxa). Our approach can help solve long-standing challenges in biodiversity research and address pressing issues in monitoring programs by saving time in sample identification.

This paper provides a bibliometric analysis of current research trends in the field of artificial intelligence (AI), focusing on key topics such as deep learning, machine learning, and security in AI. Through the lens of bibliometric analysis, we explore publications published from 2020 to 2024, using primary data from the Clarivate Analytics Web of Science Core Collection. The analysis includes the distribution of studies by year, the number of studies and citation rankings in journals, and the identification of leading countries, institutions, and authors in the field of AI research. Additionally, we investigate the distribution of studies by Web of Science categories, authors, affiliations, publication years, countries/regions, publishers, research areas, and citations per year. Key findings indicate a continued growth of interest in topics such as deep learning, machine learning, and security in AI over the past few years. We also identify leading countries and institutions active in researching this area. Awareness of data security is essential for the responsible application of AI technologies. Robust security frameworks are important to mitigate risks associated with AI integration into critical infrastructure such as healthcare and finance. Ensuring the integrity and confidentiality of data managed by AI systems is not only a technical challenge but also a societal necessity, demanding interdisciplinary collaboration and policy development. This analysis provides a deeper understanding of the current state of research in the field of AI and identifies key areas for further research and innovation. Furthermore, these findings may be valuable to practitioners and decision-makers seeking to understand current trends and innovations in AI to enhance their business processes and practices.

The black-box nature of neural networks is an obstacle to the adoption of systems based on them, mainly due to a lack of understanding and trust by end users. Providing explanations of the model’s predictions should increase trust in the system and make peculiar decisions easier to examine. In this paper, an architecture of a machine learning time series prediction system for business purchase prediction based on neural networks and enhanced with Explainable artificial intelligence (XAI) techniques is proposed. The architecture is implemented on an example of a system for predicting the following purchases for time series using Long short-term memory (LSTM) neural networks and Shapley additive explanations (SHAP) values. The developed system was evaluated with three different LSTM neural networks for predicting the next purchase day, with the most complex network producing the best results across all metrics. Explanations generated by the XAI module are provided with the prediction results to the user to allow him to understand the system’s decisions. Another benefit of the XAI module is the possibility to experiment with different prediction models and compare input feature effects.

Bioassessment is the process of using living organisms to assess the ecological health of a particular ecosystem. It typically relies on identifying specific organisms that are sensitive to changes in environmental conditions. Benthic macroinvertebrates are mostly used as bioindicators of the ecological status of freshwaters. However, a time-consuming process of species identification that requires high expertise represents one of the key obstacles to reliable bioassessment of aquatic ecosystems. Partial automation of this process using deep learning-based image classification is the goal of an ongoing project AIAQUAMI we are participating in. One of the project deliverables is a standalone desktop application for image classification with visualization and reporting that we developed and open-sourced as Imagelytics. The application relies on a convolutional neural network (CNN) to classify images and the Grad-CAM algorithm to produce heatmaps of the image areas that mostly influenced the network decision. Along with the application code, we also open-sourced scripts that can be used to train CNN on an arbitrary dataset and produce required metadata, so it can be used with Imagelytics. In this paper, we presented technical details about the application and training method that will enable its general use for image classification tasks. As a part of the evaluation, we will show a use case related to species identification of non-biting midges (Diptera: Chironomidae).

Artificial intelligence and ChatGPT became very popular in late November 2022. Their popularity has led to the development of other chat AI systems and AI tools. In order to monitor this progress and achieve key goals in the application of artificial intelligence, this paper has the task of presenting an analysis of the generated web mining program. One of the key aspects of this research is the presentation of advantages and disadvantages of this approach, including analysis of accuracy, speed and reliability. Also, potential challenges in the application of artificial intelligence (especially chat systems), such as the limitations of AI models, ethical and security aspects when processing data from the Internet, are explored. This research paper aims to contribute to a better understanding of the capabilities of AI in the development of web mining programs, identify potential areas of improvement and provide guidance for future development.

Data concerning product sales are a popular topic in time series forecasting due to their multidimensionality and wide presence in many businesses. This paper describes the research in predicting the timing and product category of the next purchase based on historical customer transaction data. Given that the dataset was acquired from a vendor of medical drugs and devices, the generic product identifier (GPI) classification system was incorporated in assigning product categories. The models built are based on recurrent neural networks (RNN) and long short-term memory (LSTM) neural networks with different input and output features, and training datasets. Experiments with various datasets were conducted and optimal network structures and types for predicting both product category and next purchase day were identified. The key contribution of this research is the process of data transformation from its original purchase transaction format into a time series of input features for next purchase prediction. With this approach, it is possible to implement a dedicated personalized marketing system for a vendor.

The use of IoT devices in the classroom has the potential to revolutionize the learning experience of students and professors. One such device is the NodeMCU, an open-source platform based on the ESP32 microcontroller. This paper presents device for registering students for classes and preparation for the experiment, which is conducted on our faculty where we work as teaching fellows, on the use of NodeMCU. In the first part of the paper, we will briefly look at some of the challenges in higher education in the Republic of Serbia and how we came up with the idea to improve the teaching process by creating a device for registering for classes. Next, we will explain in more detail the building elements of the device itself and how it works.

Recent improvements in networking technologies have led to a significant shift towards distributed cloud-based services. However, adequate management of computation resources by providers is vital to maintain the costs of operations and quality of services. A robust system is needed to forecast demand and prevent excessive resource allocations. Extensive literature review suggests that the potential of recurrent neural networks with attention mechanisms is not sufficiently explored and applied to cloud computing. To address this gap, this work proposes a methodology for forecasting load of cloud resources based on recurrent neural networks with and without attention layers. Utilized deep learning models are further optimized through hyperparameter tuning using a modified particle swarm optimization metaheuristic, which is also introduced in this work. To help models deal with complex non-stationary data sequences, the variational mode decomposition for decomposing complex series has also been utilized. The performance of this approach is compared to several state-of-the-art algorithms on a real-world cloud-load dataset. Captured performance metrics ( R^2 , mean square error, root mean square error, and index of agreement) strongly indicate that the proposed method has great potential for accurately forecasting cloud load. Further, models optimized by the introduced metaheuristic outperformed competing approaches, which was confirmed by conducted statistical validation. In addition, the best-performing forecasting model has been subjected to SHapley Additive exPlanations analysis to determine the impact each feature has on model forecasts, which could potentially be a very useful tool for cloud providers when making decisions.

Biometric security is a major emerging concern in the field of data security. In recent years, research initiatives in the field of biometrics have grown at an exponential rate. The multimodal biometric technique with enhanced accuracy and recognition rate for smart cities is still a challenging issue. This paper proposes an enhanced multimodal biometric technique for a smart city that is based on score-level fusion. Specifically, the proposed approach provides a solution to the existing challenges by providing a multimodal fusion technique with an optimized fuzzy genetic algorithm providing enhanced performance. Experiments with different biometric environments reveal significant improvements over existing strategies. The result analysis shows that the proposed approach provides better performance in terms of the false acceptance rate, false rejection rate, equal error rate, precision, recall, and accuracy. The proposed scheme provides a higher accuracy rate of 99.88% and a lower equal error rate of 0.18%. The vital part of this approach is the inclusion of a fuzzy strategy with soft computing techniques known as an optimized fuzzy genetic algorithm.

The analysis of community structure in studies of freshwater ecology often requires the application of dimensionality reduction to process multivariate data. A high number of dimensions (number of taxa/environmental parameters × number of samples), nonlinear relationships, outliers, and high variability usually hinder the visualization and interpretation of multivariate datasets. Here, we proposed a new statistical design using Uniform Manifold Approximation and Projection (UMAP), and community partitioning using Louvain algorithms, to ordinate and classify the structure of aquatic biota in two-dimensional space. We present this approach with a demonstration of five previously published datasets for diatoms, macrophytes, chironomids (larval and subfossil), and fish. Principal Component Analysis (PCA) and Ward's clustering were also used to assess the comparability of the UMAP approach compared to traditional approaches for ordination and classification. The ordination of sampling sites in 2-dimensional space showed a much denser, and easier to interpret, grouping using the UMAP approach in comparison to PCA. The classification of community structure using the Louvain algorithm in UMAP ordinal space showed a high classification strength for data with a high number of dimensions than the cluster patterns obtained with the use of a Ward's algorithm in PCA. Environmental gradients, presented via heat maps, were overlayed with the ordination patterns of aquatic communities, confirming that the ordinations obtained by UMAP were ecologically meaningful. This is the first study that has applied a UMAP approach with classification using Louvain algorithms on ecological datasets. We show that the performance of local and global structures, as well as the number of clusters determined by the algorithm, make this approach more powerful than traditional approaches.