Human Activity Recognition (HAR) datasets contain complex patterns that supervised models exploit with labeled training, but it remains unclear what latent structure exists in the data itself. This paper presents an unsupervised exploratory analysis of a smartphone sensor HAR dataset to uncover inherent activity groupings without using activity labels. We apply a range of clustering algorithms (k-means, Gaussian mixture, hierarchical agglomerative, density-based HDBSCAN, spectral clustering) and dimensionality reduction methods (Principal Component Analysis – PCA, t-distributed Stochastic Neighbor Embedding – t-SNE, Uniform Manifold Approximation and Projection – UMAP, and a feed-forward autoencoder) to identify natural clusters of sensor feature vectors. Quantitatively, we evaluate clustering quality using internal metrics (silhouette coefficient) and external metrics against true labels (Adjusted Rand Index – ARI, and Normalized Mutual Information – NMI). The results reveal a dominant two-cluster division separating static postures from dynamic movements, with finer sub-clusters roughly corresponding to the six known activities when clustering is applied hierarchically. UMAP non-linear embedding dramatically improved cluster separability and alignment with classes, outperforming PCA. Analyzing feature importance in each cluster showed that features related to body orientation and acceleration dynamics differentiate activities. These findings demonstrate that unsupervised learning can automatically discover meaningful activity groupings (e.g. distinguishing stationary vs. moving behaviors) and key distinguishing sensor features, without any labels. The study provides insights into intrinsic HAR data structure, which can inform feature design and hierarchical modeling in future activity recognition systems.

YOLO-based object detectors support fast wildlife monitoring across diverse habitats. Static training limits these detectors in the field because ecological conditions shift over time. Continual learning provides an incremental update process that preserves earlier knowledge while absorbing new information. This article reviews recent progress in continual learning for YOLO detection, with emphasis on experience replay, self-distillation, and human-in-the-loop supervision. These approaches protect past knowledge, reduce annotation demands, and deliver more stable predictions during long-term deployments. A focused case study on Greater One-Horned Rhino monitoring shows how adaptive learning pipelines strengthen detection reliability during seasonal changes, altered terrain, and new camera placements. The review outlines methods with strong potential for long-term conservation work and highlights future directions for resilient wildlife monitoring systems.

The increasing world demand for eco-friendly natural fibers and products is gaining interest in sustainable fibers as alternatives to man-made fibers. Bast fibers such as linen, banana, jute, bamboo, sisal, kenaf, and Mesta have been used for a long time. Bast fibers are the raw materials of the present and the future, not just for textiles but also for modern eco-friendly composite materials, medicine, cosmetics, food, biopolymers, agro-fine chemicals, and energy. The retting is the biggest problem with extracting the fibers. Dew retting as well as stagnant water and running water retting are the conventional methods to separate the long bast fibers. Due to higher tensile property, mechanical performance, environmental benefits, and economic viability, Mesta has shown significant promise compared to synthetics. They can be grown in numerous climates, and, under the right conditions, they don't hurt the ecology too much or at all. The study looks into the important physical and mechanical properties and benefits of Mesta fibers, as well as their possible uses. It focuses on how they might help promote sustainable consumption and green industrial practices. The review talks on Mesta's environmental, technical, and social-economic importance, showing that they are still crucial for accomplishing ecologically sustainable development goals and coming up with novel bio-based materials.

This systematic review explores the selection and utilization of sustainable materials in building design and construction, emphasizing their environmental, economic, and social impacts. The review follows the PRISMA guidelines, identifying 50 relevant studies published between 2010 and 2023. The findings highlight that sustainable materials, including recycled steel, bamboo, and low-carbon concrete, significantly reduce greenhouse gas emissions, energy consumption, and resource depletion compared to traditional materials. Life Cycle Assessment (LCA) proved crucial in evaluating these environmental benefits. Economically, although the initial costs of sustainable materials are often higher, their long-term financial advantages—such as reduced operational costs, energy savings, and lower maintenance expenses—make them viable investments. Market trends indicate that growing demand is gradually lowering the costs of these materials. Socially, sustainable materials improve indoor air quality, reduce the health risks associated with volatile organic compounds (VOCs), and enhance occupant well-being, promoting community engagement by supporting local economies. Despite these benefits, challenges remain, particularly regarding the availability and cost of sustainable materials in developing regions. The review concludes that overcoming these barriers requires continued technological advancements, government incentives, and more robust regulatory frameworks to accelerate the adoption of sustainable building practices. Overall, the review emphasizes the critical role of sustainable materials in addressing climate change, promoting economic sustainability, and fostering social inclusivity in construction while underscoring the need for global efforts to support the transition towards eco-friendly and resilient built environments.

This study investigates applying advanced machine learning techniques in enhancing cybersecurity systems, particularly in phishing detection, network intrusion detection, and malware and ransomware classification. Supervised learning algorithms such as random forests and support vector machines (SVM), deep learning models like convolutional neural networks (CNN) and recurrent neural networks (RNN), and ensemble methods were employed to improve detection accuracy and reduce false positives. The study also addresses key challenges, including adversarial attacks, data imbalance, and the need for continuous learning to adapt to evolving threats. Results indicated that machine learning models, especially deep learning techniques, demonstrated high accuracy in detecting complex threats, with phishing detection models achieving over 96% accuracy and network intrusion detection models reaching 98.2%. The study also explored the use of transfer learning and continuous learning systems, which showed promise in adapting to new threats while minimising the need for extensive retraining. However, adversarial vulnerabilities and the challenge of catastrophic forgetting in continuous learning models remain significant obstacles. Recommendations include integrating adversarial training, improving data augmentation techniques, and optimising continuous learning systems for real-time threat adaptation. This research contributes to the growing body of knowledge on machine learning applications in cybersecurity, highlighting both its potential and the need for ongoing refinement to address emerging cyber threats.

This study systematically reviews the integration of machine learning (ML) and artificial intelligence (AI) into SQL databases and big data analytics, highlighting significant advancements and emerging trends. Using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, a comprehensive review of 60 selected articles published between 2010 and 2023 was conducted. The findings reveal substantial improvements in query optimization through ML algorithms, which adapt dynamically to changing data patterns, reducing processing times and enhancing performance. Additionally, embedding ML models within SQL databases facilitates real-time predictive analytics, streamlining workflows, and improving the accuracy and speed of predictions. AI-driven security systems provide proactive and real-time threat detection, significantly enhancing data protection. The development of hybrid systems that combine relational and non-relational databases offers versatile and efficient data management solutions, addressing the limitations of traditional systems. This study confirms the evolving role of AI and ML in transforming data management practices and aligns with and extends previous research findings.

Given the rising demand for big data processing and the current dynamic economic environment, which presents challenges for businesses, especially in terms of managing cloud computing costs has become a large issue. Although cloud infrastructure provides scalability and adaptability, poor management of it could cause significant expenses. Large data set processing inside distributed computing systems largely relies on Amazon Elastic MapReduce (EMR). Although Electronic Medical Records in the healthcare sector and other businesses managing large amounts of data might find EMS suitable, its dynamic and scalable capabilities could potentially lead to cost inefficiencies. Inappropriate scaling or too generous resource allocation can lead to resource waste and higher running expenses. This white paper examines cutting-edge cost-optimal methods designed especially for E MR workload management. Electronic Medical Records (EMRs) are confidential patient records requiring efficient, safe, reliable processing—qualities lacking in which case significant charges could arise. The study stresses on a simple approach the usage of tailored application programming interfaces (APIs). These APIs let one automate important chores such dynamic job scheduling, real-time instance selection, autonomous scaling, and ongoing cost monitoring. By means of automation, businesses may guarantee that computing resources are distributed precisely where and when needed, therefore avoiding the inefficiencies connected with set configurations. Dynamic work scheduling distributes tasks depending on real-time data, therefore optimizing resource use all day. Companies can identify the most reasonably priced and task-appropriate computer instances by means of selective instance selection, therefore avoiding a homogeneous approach that might result in inefficiencies or insufficient performance. By allowing systems to dynamically change resources in response to various demands, autonomous scaling guarantees performance while eliminating needless resource allocation and so maximizes efficiency.

Radio frequency (RF) and electromagnetic (EM) systems still rely heavily on half-wave dipole antennas due to their practical design, reliable operation, and adaptability to a variety of communication applications. The development of contemporary wireless technologies such as 5G, the Internet of Things (IoT), and wearable technology has led to a growing demand for accurate dipole antenna design and analysis to fulfill performance requirements. Because simulation tools may save development cycles and improve the accuracy of design forecasts, RF engineers are increasingly using them to do this. Of them, MATLAB has emerged as a popular platform for antenna modeling because it provides a very flexible environment that accommodates both sophisticated, algorithm-level modifications and streamlined design procedures. This article provides a thorough analysis of MATLAB-based methods for half-wave dipole antenna modeling. Antenna Toolbox, user-developed solutions using the Method of Moments (MoM), and hybrid approaches that combine MATLAB with full-wave solvers like CST Microwave Studio and Ansys HFSS are all covered. Each method is evaluated based on its precision, processing requirements, flexibility, and applicability to certain use cases. The balance between usability, level of control over the modeling process, and the accuracy of the simulated outcomes is among the trade-offs included in the analysis. The study discusses the shortcomings of these approaches as well as their advantages, including the lack of generally recognized benchmarking standards, the abstraction level of solvers, and the low precision of nearfield modeling. This review is useful for both academic researchers and professionals in the business since it provides a comprehensive assessment of existing procedures and identifies areas that require improvement. It attempts to guide anyone looking for dependable and effective MATLAB antenna modeling solutions, as well as suggestions for future advancements in simulation-driven antenna design.