Gongcheng Kexue Yu Jishu/Advanced Engineering Science

The aggressive scaling of CMOS technologies and the widespread adoption of Internet of Things (IoT) systems demand highly power-efficient, temperature-robust, and frequency-stable on-chip power management and clock generation circuits. This work presents an integrated design of a low-power bandgap reference (BGR), a temperature-adaptive low-dropout regulator (LDO), and a low power crystal oscillator to achieve reliable timing under wide operating conditions. Finally, a low power crystal oscillator is implemented using a Pierce configuration with negative resistance of approximately 500Ω at 48MHz, satisfying the oscillation startup criterion of three times the crystal ESR. The oscillator achieves rail-to-rail output swing, startup time below 2ms, and total current consumption below 75µA, corresponding to power consumption of approximately 90µW at 1.2V. The combined architecture demonstrates that co-design of reference, regulation, and clock generation circuits significantly enhances frequency stability, reduces power dissipation, and improves system-level power efficiency, making the proposed framework highly suitable for next-generation IoT, wearable, and battery-operated edge devices.

Molecular structures are assigned numerical v-alues known as topological indices, which serve as important descriptors in chemical graph theory. In the present study, degree-based topological indices of the total graph of alkanes, a significant class of saturated hydrocarbons, are investigated. Several widely used indices, including the first and second Zagreb indices, Randić index, harmonic index, atom-bond connectivity index, forgotten index, sum connectivity index, symmetric division degree index, geometric–arithmetic index, and the first and second hyper Zagreb indices, are computed for the total graphs of linear alkanes. A detailed numerical comparison of these indices is presented along with their graphical behaviour as the carbon chain increases. Furthermore, regression analysis is performed to examine the relationship between the computed indices and selected physicochemical properties. The results highlight the influence of molecular structure on topological characteristics and demonstrate the applicability of total graph of alkanes-based descriptors in chemical graph theory, molecular modeling, and predictive chemistry.

Graph operations are essential in many graph theory applications because they enable the creation of large graphs from smaller ones. This study focuses on a particular graph-theoretical operation: the Cartesian product graph. Our research is focused on alkanes, the simplest hydrocarbons made only of hydrogen (H) and carbon (C) atoms with no functional groups. In this paper, we compute several topological indices for the Cartesian product of alkanes namely, Atom Bond Connectivity Index ( ABC), Geometric Arithmetic ( GA ), Randic Index ( ), Reciprocal Randic Index ( RR ), First Zagreb Index ( ), Second Zagreb Index ( ), Hyper Zagreb Index ( HM ), Forgotten Index ( F ), Inverse Sum Indeg Index ( ISI ), Symmetric Division Deg Index ( SDD ). Additionally, we provide a graphical and numerical comparison of these topological indices.

The rapid growth of IoT (Internet of Things) networks has brought new challenges in ensuring the safety and reliability of devices and data. Automated Teller Machines (ATMs) are increasingly targeted by physical attacks, theft, and sophisticated intrusion attempts, necessitating intelligent and privacy-preserving security mechanisms. This paper proposes a federated learning–based multi-level threat detection framework for ATM security that integrates both external sensors (vibration, magnetic, tilt) and internal sensors (door status, liveness detection, face masks or static images). Each ATM operates as an edge node, locally analysing sensor data to classify operational states into three threat levels: Normal, Theft, and Critical. To ensure reliability and safety, a hybrid decision strategy is adopted, combining rule-based safety checks for critical events with a simple machine learning model for nuanced threat identification. Collaborative model training across several ATMs is made possible by federated learning. without transferring raw sensor data to a central server, protecting data in the process and reducing communication overhead. Critical conditions such as forced entry attempts, abnormal cash handling, and ATM communication failures are detected in real time and escalated through centralized monitoring using visual alert or notify the bank’s higher authorities. The findings show that federated learning is a scalable and secure solution for distributed ATM threat monitoring in real-world banking infrastructures, maintaining data privacy, reducing communication overhead.

Healthcare systems face an unprecedented surge in sophisticated cyberattacks, with ransomware, data breaches, and advanced persistent threats (APTs) posing severe risks to patient safety and data privacy. Traditional security information and event management (SIEM) systems struggle to keep pace with the evolving threat landscape, demonstrating detection rates below 75% for novel attacks and generating excessive false positives that overwhelm security operations centers. This paper presents a comprehensive AI-driven cybersecurity framework specifically designed for healthcare environments. Our approach integrates multiple machine learning algorithms including Isolation Forest for anomaly detection, Long Short-Term Memory (LSTM) networks for temporal pattern recognition, and ensemble methods combining Random Forest, Convolutional Neural Networks (CNN), and Deep Neural Networks (DNN) to achieve superior threat detection and automated response capabilities. Through extensive evaluation across simulated and real-world healthcare network environments, our ensemble model achieved 98.5% accuracy in threat detection, with precision of 98.2%, recall of 97.9%, and an F1-score of 98.0%. The false positive rate was reduced to 1.8%, compared to 12.5% for traditional SIEM systems. Average threat detection time improved from 45-240 minutes (traditional methods) to 1.8-5.2 minutes (AI-enhanced system), enabling rapid response to critical threats. The framework incorporates automated response mechanisms, continuous learning capabilities, and HIPAA-compliant data handling procedures, making it practical for deployment in resource-constrained healthcare environments. Implementation costs range from $400K-$800K, with projected ROI of 220-340% within 24 months through reduced breach incidents, minimized downtime, and lower operational overhead.

Effective healthcare delivery and research increasingly depend on the ability to share clinical data across organizational boundaries. However, fragmented systems, escalating cyber threats, and stringent regulatory requirements continue to limit secure and trustworthy health information exchange. Traditional perimeter-based security models, which assume implicit trust within institutional networks, are no longer adequate in modern, highly distributed healthcare environments. This paper presents a comprehensive framework for secure health data sharing grounded in continuous verification principles, layered encryption, and coordinated governance across participating institutions. The proposed architecture enforces verification of every access request regardless of user location, encrypts data throughout its lifecycle, and maintains detailed audit trails to ensure accountability and regulatory compliance. In addition, the framework supports privacy-preserving collaborative research through federated and distributed analysis models that allow institutions to generate shared insights without exposing individual patient records. Practical implementation strategies, governance structures, and phased deployment pathways are outlined to enable incremental adoption without disrupting clinical workflows. By integrating technical safeguards with organizational coordination, this framework demonstrates how healthcare organizations can balance data accessibility, patient privacy, and system security. The approach offers a scalable and practical foundation for advancing interoperable care, collaborative clinical intelligence, and public trust in health data exchange.

The increasing use of multi-layer composite materials in high-performance engineering applications has created a growing demand for accurate modeling of transient heat conduction. The heterogeneous and anisotropic nature of composite systems, combined with the presence of multiple material interfaces, introduces significant complexity in predicting time-dependent thermal behavior. This study presents a qualitative and systematic synthesis of existing research on transient heat conduction modeling in multi-layer composite materials with the objective of developing an integrated conceptual framework. A systematic literature review was conducted using major scientific databases, and the selected studies were analyzed using thematic analysis to identify dominant modeling approaches, key influencing parameters, and persistent research challenges. The findings reveal the predominance of numerical techniques, particularly finite element modeling, supported by analytical methods for theoretical validation. The analysis highlights the critical role of thermal interface resistance, anisotropic material behavior, and the increasing importance of multi-scale and multi-physics modeling. Emerging trends such as machine learning and digital twin technologies were identified as promising directions for improving predictive accuracy and reducing computational cost. The study identifies significant gaps in experimental validation, interface characterization, and integration of physics-based and data-driven approaches. Based on the synthesis, an integrated conceptual framework for transient heat conduction modeling is proposed. The findings contribute to a more unified understanding of thermal behavior in layered composite systems and provide guidance for future research in advanced thermal management and composite material design.

Detection of images in low-visibility poses challenges for safety management of autonomous vessels, coastal surveillance, and maritime situational awareness. Our emotional convolution neural network suffers from performance degradation when raw images are captured under adverse weather, fog and in low-light (dark) illumination. We propose a light-weight Vision Transformer (ViT) based model which provides real-time nautical image enhancement and object detection for lower visibility conditions. We propose a paradigm shift from using a convolution neural network (CNN) as the work-horse of ViTs are good models for capturing long range information but are computationally light-weight and amenable to low-Size Area Power (SAP) embedded systems space applications. Having surveyed and re-used some existing light-weight magic tricks we integrate three ingredients: (1) - a contrast-guided image augmentation module on the basis of dark channel prior, or non-uniform illumination correction techniques [essentially re-establishing “daylight” illumination], (2) a light-weight vision transformer backbone based on Rostrum Rep ViT allowing us to utilize a proportionately lower number of parameters along with the beauty of transformers, and (3) a scale-adaptive feature fusion module for distributing information through cross stage connections. We provide inner workings Anglo-Saxon Esque explanations plus expectations in terms of increasing transparency to underpin the advantageous use of our cherries here. Furthermore, recipe ingredient/Ton berry poison testing using ablation studies confirms miraculously that each of our features/components have “magical” merit. We perform experiments in several data sets RUOD, UTDAC etc and also low-visibility maritime situations (including dense fog). Testing shows a mean Average Precision (mAP) of 94.7% on 321, LOL380 speeds on edge systems via the Py Torch framework. Our model has direct applicability in dense fog and heavy rain amongst other tests, while we show our method achieves required performance with a truly marginal added computational increase 12.8M parameters 18.4 GFLOPs compared to a number of other CANN solutions including YOLOv11 based and RT-DETR. Our work closes the crucial gap in maritime automation that now exists, providing a simple, deployable method to enable real-time object detection on low-resource platforms without sacrificing accuracy, directly impacting autonomous surface vessel (ASV) navigation, maritime port security, fishing vessel monitoring, and climate resilient coastal management systems in keeping with the IMO 2030/2050 decarbonisation targets.

The emergence of artificial intelligence in textual production marks a profound transformation in the structure of digital literature and cultural expression. Unlike earlier digital innovations that primarily altered the format or distribution of texts, AI-driven systems actively participate in generating language itself. Large language models produce coherent prose, poetry, and narrative structures through probabilistic pattern recognition, drawing from vast datasets of human-authored material. As a result, textual creation shifts from intentional authorship toward computational synthesis, challenging long-standing assumptions about originality, creativity, and meaning. Traditionally, literary meaning has been associated with human experience, authorial intention, and interpretive context. However, AI-generated texts complicate this framework by demonstrating linguistic fluency without consciousness or subjective awareness. This development raises critical theoretical questions: If language can be produced without lived experience, does meaning remain tied to intention, or does it emerge entirely through reader interpretation? In the digital environment, meaning increasingly becomes relational and interactive, shaped by algorithmic generation, platform circulation, and audience engagement. At the same time, artificial intelligence operates within broader cultural infrastructures that influence which texts are produced, amplified, and consumed. Algorithmic systems not only generate language but also rank, recommend, and distribute content, thereby shaping cultural visibility and symbolic power. Consequently, digital texts in the age of AI reflect both creative possibility and structural constraint. The transformation of literary culture under artificial intelligence is therefore not merely technological but epistemological. It redefines language as data-driven synthesis and repositions cultural expression within computational systems, prompting a re-evaluation of authorship, creativity, and the nature of literary meaning.

A novel polymer was synthesized through the condensation of 2, 4-dihydroxybenzoic acid with acrylamide and formaldehyde. The polymer was comprehensively characterized by elemental analysis, physico-chemical methods, and spectrometry such as Ultraviolet Visible spectroscopy, Infrared spectroscopy and Proton Nuclear Magnetic Resonance. Semi-crystalline nature of polymer was exported using electron Beam microscopy. GPC (gel permeation chromatography) was implemented to figure out the molecular weight of the polymer. To assess its metal adsorption capabilities, a batch separation performance was used for the selective extraction of toxic traces metallic cation such as Co²⁺, Cd²⁺, and Pb²⁺. Experimental outcomes indicated that the synthesized polymer possesses a highly permeable and major surface area, contributing to its enhanced metal ion uptake. When differentiate with marketable available phenolic and polystyrene resins, the polymer demonstrated superior adsorption performance. By employing thermogravimetric analysis (TGA), the polymer's thermal stability has been assessed. Detailed thermal decomposition behavior was studied, and kinetic parameters were evaluated using both the Freeman–Carroll and Sharp–Wentworth methods. The consistency between the results from both methods confirmed the reliability of the thermal analysis.

Narrative voice functions as a critical medium through which identity is constructed, negotiated, and reconstructed in contemporary English literature. In contexts shaped by globalization, migration, postcolonial legacies, gender politics, and socio-cultural fragmentation, literary narratives increasingly foreground voice as a site of memory, resistance, and power negotiation. This study examines how contemporary English literature reconstructs identity through linguistic strategies, narrative perspective, and memory articulation. Drawing from postcolonial theory, feminist narratology, and discourse analysis, the paper argues that narrative voice is not merely a stylistic device but a political and epistemological instrument that shapes subjectivity and challenges dominant power structures. By analyzing selected thematic patterns in contemporary texts, the study explores how fragmented narration, multilingual expression, unreliable narrators, and memory-based storytelling contribute to identity formation. The paper concludes that narrative voice serves as a transformative space where marginalized identities reclaim agency and contest hegemonic discourses.

The impact of increased CO2 in environment is a major factor of climate change and has cascading effects on ecosystem, human societies and agriculture. Increased CO2 levels can affect crop yield and food security. This paper aims to examine carbon credit of a greenhouse integrated solar photovoltaic thermal (GiSPVT) system that uses Multi crystalline solar cell material with semitransparent (SPVT) collectors for control environment while promoting the friendly atmosphere. GiSPVT-integrated air collectors supply thermal and electrical energy, which is used to regulate the structure's humidity and temperature. In this paper analytical calculations are done to derive the expressions that can find temperature of the plant and air within the greenhouse, solar cell temperature, also the efficiency of electrical energy of the GiSPVT system along with the collectors. The expressions are derived with the help of climatic variables like solar irradiations and ambient air temperature, as well as some design parameters such as temperature coefficients, heat transfer coefficients, area of the PV module, electrical efficiency under standard test conditions (STC). The performance of the proposed system is evaluated using different energy performance indicators, namely energy payback time (EPBT), energy production factor (EPF), and life cycle conversion efficiency (LCCE). Thermal modelling is specifically carried out for a multi-crystalline silicon (mc-Si)–based GiSPVT system integrated with air collectors. The total electrical and thermal energy outputs of the system are estimated in order to assess these energy matrices. In addition, the total carbon emissions, carbon mitigation, and carbon credits associated with the proposed system over its entire lifetime are also evaluated.

The stagnation of commercial productivity in okra [Abelmoschus esculentus (L.) Moench] due to narrow genetic bases and climatic fluctuations necessitates a paradigm shift from mere yield maximization to the development of "Climate Smart" genotypes. Understanding Genotype × Environment (G × E) interactions are critical to identifying hybrids that maintain stability across diverse agro-climatic zones.The present investigation evaluated the phenotypic stability and adaptability of 37 genotypes (8 diverse parents, 28 F1 hybrids derived from a half-diallel mating design, and one standard check, Parbhani Kranti). The experiment was conducted across three distinct environments (E1, E2, E3) using a Randomised Block Design. Stability parameters were estimated using the Eberhart and Russell (1966) model to partition G × E interaction into linear (predictable) and non-linear (unpredictable) components for eleven quantitative traits. Results: Pooled analysis of variance revealed highly significant differences among genotypes and significant G × E interactions for key economic traits, including fruit yield per plant. Notably, the linear component of G × E interaction was significant for fruit yield, indicating that the performance shifts of these hybrids are predictable. The stability analysis identified Phule Utkarsha × AKOV-107 as a specific adaptor for favourable, high-input environments, recording the highest mean yield of 163.65 g/plant with a regression coefficient (bi) of 2.04. Conversely, the hybrid AKOV-107 × GO-6 emerged as a widely adapted, resilient genotype with a stable mean yield of 118.90 g/plant and a regression coefficient near unity (bi ≈ 0.82). For resource-constrained conditions, GOA-5 × GJO-3 demonstrated stress tolerance with a distinct stability profile (bi < 0). Conclusion: The study successfully identified novel heterotic combinations that break the current yield plateau while offering specific adaptation strategies. The identification of these stable and responsive hybrids provides a roadmap for "Climate Smart" okra breeding, ensuring nutritional security under fluctuating climatic conditions.

IoV is becoming a major facilitator of intelligent transport systems, which can encompass vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), and vehicle-to-everything (V2X) communications. Nonetheless, IoV networks can be susceptible to fake information transmission, Sybil attacks, and malicious nodes that decrease the reliability of vehicle exchange of data. The current centralized trust authorities have a single point of failure, bottlenecks in scalability, and poor transparency. In response to those challenges, the paper suggests blockchain-based decentralized trust and reputation management system of the IoV. The system proposed combines a lightweight blockchain consensus among Road Side Units (RSUs) and edge servers with a smart contract-based trust model to assign reputation scores to vehicles dynamically, depending on the accuracy, consistency, and behavior of the data. The blockchain holds trust scores that are imperishable and enable reputation management that is both transparent and incorruptible. The framework is feasible as evidenced by a simulated environment of synthetic vehicular transactions Simulation results indicate faster trust convergence, effective malicious vehicle discrimination, and scalability trends under increasing vehicular density. The findings are based on comparative and theoretical examination, which provides a base to further implementation based on SUMO and Hyperledger Fabric. This study offers a safe and scalable solution to trust management in IoV and opens the gate to safer and more trusted intelligent transport systems.

Traditional artificial intelligence (AI)–based inspection systems in manufacturing are typically deployed as static classifiers, trained on historical datasets and rarely updated to reflect evolving production conditions. However, contemporary manufacturing environments are characterized by concept drift, rare defect emergence, process variability, and increasing regulatory scrutiny, rendering static inspection architectures inadequate. This study proposes a socio-technical framework of Adaptive Artificial Intelligence (AAI) for industrial quality inspection. We conceptualize AAI as an enterprise-level dynamic capability that integrates adaptive learning mechanisms, uncertainty-aware decision engines, dynamic inspection policies, human-in-the-loop governance, and enterprise system alignment into a unified architecture. Grounded in adaptive systems theory, dynamic capabilities theory, socio-technical systems research, and cyber-physical systems literature, the framework formalizes five interrelated constructs and articulates their structural relationships through moderated and mediated theoretical propositions. Specifically, adaptive learning moderates the relationship between production volatility and inspection robustness, while dynamic inspection policy mediates the impact of uncertainty-aware decision signals on cost–quality efficiency outcomes. Governance and enterprise alignment act as cross-cutting enablers influencing organizational trust and learning. The paper further redefines inspection performance through resilience-oriented metrics, including robustness under drift, adaptive capacity, inspection elasticity, cost–quality efficiency, and trust-based compliance. Cross-industry illustrations from automotive, electronics, and pharmaceutical manufacturing demonstrate the applicability of the framework. By repositioning inspection as a dynamic, enterprise-embedded capability rather than a static algorithmic function, this study contributes to operations and AI research and advances the theoretical foundations of human-centric Industry 5.0 quality systems.

Financial literacy is extremely significant in today's complex and interconnected global environment. It encompasses the knowledge, skills, and understanding of financial concepts that empower individuals to make informed and responsible decisions about their personal finances. The present study focuses on determining the financial literacy among students at higher educational institutes, the challenges they encounter in managing their personal finances, and explores the measures that can be taken to improve their financial literacy and overall financial well-being. Using both main and secondary sources, a survey-based research design gathered data from 101 college students. Participants were divided into low, medium, and high literacy groups based on self-evaluation and knowledge of fundamental financial concepts. The results indicate that 20.79% of students exhibit low literacy, 28.71% exhibit high literacy, and 50.5% exhibit moderate financial literacy. While only a small fraction turns to professional financial counsellors, most students (44.6%) depend on self-directed research for financial knowledge. Lack of financial knowledge (25.74%), peer pressure causing overspending (23.76%), and high cost of living (20.79%) were among the main issues discovered. Interestingly, 68.32% of pupils advocate making financial literacy instruction required in every field of study. The research supports the conclusion that better personal financial management is strongly correlated with higher financial literacy and that financial education initiatives successfully raise pupils' financial awareness. But financial literacy levels were not much different between various academic fields. The study advises including financial literacy in academic programs, holding frequent seminars, and using digital tools to help pupils improve their financial capacity. For governments trying to enhance financial education programs for college students as well as for educational institutions, these results offer insightful information.

In this paper, we introduce the concept of neutrosophic soft regular semi baire spaces in neutrosophic soft topological spaces. Also, we define neutrosophic soft regular semi-nowhere dense, neutrosophic soft regular semi first category and neutrosophic soft regular semi second category sets. Some of its characterizations of neutrosophic soft regular semi baire spaces are also studied. Several examples are given to illustrate the concepts.

Dehydration and electrolyte derangement constitute critical physiological stressors, particularly in environments characterized by extreme thermal indices and high physical exertion. Despite the clinical significance, continuous monitoring of hydration status remains a challenge due to the lack of non-invasive, cost-effective, and integrated real-time tracking systems. This paper proposes a conceptual Internet of Things (IoT) framework designed for autonomous physiological and environmental telemetry. The proposed architecture leverages an ESP32 microcontroller as a central processing unit, integrating high-fidelity sensors for heart rate (HR), body temperature, ambient humidity, and atmospheric pressure. A novel mathematical Hydration Index (HI) model is developed using min-max normalization and a weighted multi-sensor data fusion algorithm. Unlike traditional reactive measures, this framework proactively estimates fluid-electrolyte status by quantifying the interplay between internal cardiovascular stress and external environmental variables. Initial simulation-based analysis demonstrates the model's logical consistency in differentiating between normal hydration, mild dehydration, and acute risk states. The system architecture emphasizes a layered IoT approach, ensuring high scalability and low-power consumption for integration with mobile health (mHealth) platforms. This research provides a structured foundation for non-invasive health supervision in sports science, geriatric care, and occupational safety. Future extensions will focus on longitudinal experimental validation and the integration of predictive machine learning models for personalized hydration risk assessment.

In recent years, speech recognition technology has gained significant importance in web-based applications due to its ability to enable hands-free interaction, enhance user convenience, and improve accessibility for individuals with physical or hearing impairments. The rapid advancement of artificial intelligence and natural language processing techniques has made voice-driven systems more accurate, responsive, and adaptable across multiple domains such as education, healthcare, customer service, and smart environments. Voice-to-text conversion, in particular, has become a widely adopted feature in modern applications, allowing spoken language to be transformed into written text efficiently and in real time. This article presents the design and implementation of a Voice-to-Text Based Threat Detection and Alert System developed using Python and the Django web framework. The system is designed as a web-based application that captures spoken input directly through a browser interface, processes the audio using the SpeechRecognition library, and converts it into readable textual format. The transcribed text is displayed instantly on the application interface, ensuring smooth user interaction and minimal processing delay. By leveraging Django’s structured backend architecture, the system ensures secure data handling, efficient request processing, and scalable deployment capabilities. Unlike conventional voice-to-text systems that focus solely on transcription accuracy and speed, this system extends functionality by incorporating an intelligent threat detection module. After the speech is converted into text, the system analyzes the content using predefined keywords, logical filtering rules, and pattern recognition techniques to identify potentially harmful, abusive, or threatening language. This additional layer of analysis transforms the system from a simple transcription tool into a proactive safety monitoring solution. When suspicious or dangerous content is detected, the system automatically triggers an alert mechanism. A notification is generated and displayed on a dedicated monitoring dashboard designed for parents, guardians, or administrators. The alert includes relevant details such as the identified keyword and transcript segment, enabling timely awareness and intervention. All conversation logs and alert records are securely stored in the database for documentation and future review. By combining real-time speech processing, web technologies, structured data management, and safety monitoring mechanisms, the proposed system offers an efficient, scalable, and user-friendly solution for secure communication environments. The integration of voice recognition with automated threat detection demonstrates how modern technologies can be utilized not only to enhance accessibility and convenience but also to strengthen digital safety and proactive supervision.

Crop diseases are a major threat to agricultural productivity and food security worldwide, directly impacting the achievement of SDG 2 – Zero Hunger established by the United Nations. Early detection of plant diseases is crucial to prevent yield loss, enhance food availability, and ensure sustainable farming practices. Traditional methods of identifying crop diseases rely on manual inspection by experts, which can be time-consuming, costly, and prone to human error. This project presents an intelligent system for early detection of crop diseases using image processing techniques and Convolutional Neural Networks (CNN). The proposed method begins with capturing high-resolution images of crop leaves using a mobile camera. The images undergo pre-processing steps such as noise removal, resizing, and contrast enhancement to improve image quality. After pre-processing, image segmentation is performed to isolate the diseased region from the healthy portion of the leaf. Feature extraction is automatically handled by the CNN model, which identifies patterns related to color variations, texture irregularities, and shape distortions. The extracted features are passed into the trained CNN classifier, which categorizes the leaf as healthy or affected by a specific disease. The system provides accurate and real-time disease prediction. By enabling early intervention, this approach reduces dependency on manual diagnosis, minimizes crop loss, increases agricultural productivity, and strengthens food security. Therefore, the proposed system directly contributes to achieving SDG 2 – Zero Hunger by supporting sustainable agriculture and improving global food production through precision farming technologies.

Detecting chronic diseases at an early stage is a major contributing factor to reducing healthcare expenses and increasing survival rates. Most diagnostic methods utilize laboratory tests and the physician’s expertise; therefore, it could take time for patients to receive timely treatment.This research presents a Multi-Disease Prediction and Appointment Management System that employs machine learning techniques in order to create an integrated solution for providing healthcare. The proposed system will predict four major diseases, including: diabetes, heart disease, Parkinson’s disease, and skin diseases. Specifically, Support Vector Machine (SVM) will be used to predict diabetes and Parkinson’s Disease, Logistic Regression will be used to predict heart disease, and Convolutional Neural Networks (CNN) will be used to categorize dermatological images. Medical datasets and images of skin conditions were collected from Kaggle; the datasets were then trained and tested in Google Colab. After training and testing the model in Google Colab, it was deployed via a web interface created with Streamlit. Additionally, Firebase will be utilized as a mechanism for both secure doctor/patient authentication and real-time appointment management. This study contributes to sustainable healthcare development by enhancing public health, improving health system access, and supporting inclusive health through machine learning–based disease prediction.

Electrical cable theft is a major problem in Indian Railways. It leads to service interruptions, financial losses, and serious safety risks. Traditional monitoring systems rely on manual checks or limited CCTV surveillance. They frequently fail to catch theft in real time, causing delays in response and increased damage. This project presents an Electrical Cable Theft Detection System using IoT sensors, microcontrollers, and GSM communication modules to monitor cable continuity and voltage levels along railway tracks. The system promptly detects unusual changes like cable cuts or voltage drops. When it identifies theft, it sends immediate alerts to railway authorities, enabling quick actions to reduce damage. By providing real-time monitoring, the system enhances the security of railway electrical infrastructure, lowers financial losses due to theft or damage, and ensures passenger safety. Moreover, it decreases the need for manual inspections and lessens manpower requirements. It supports continuous monitoring even in remote or hard-to-reach railway areas. Overall, this IoT-based solution offers a reliable, effective, and affordable way to safeguard crucial railway infrastructure

New trends in power system includes the optimal placement of Flexible AC transmission System (FACTS) devices to improve the performance of the system by managing the power by managing the load. A meta heuristic method is proposed for the optimal placement and sizing of FACTS devices, namely the Thyristor-Controlled Series Compensators (TCSCs), Shunt VARs Compensators (SVCs), and Unified Power Flows Controllers (UPFCs). To find the optimal locations of these devices in a network, weak buses and lines are determined through the line sensitivity index, Then, Hybrid GA-PSO is employed not only to find an ideal rating for these devices but also the optimal coordination of SVC, TCSC, and UPFC with the reactive power sources already present in the network (tap settings of transformers and reactive power from generators). The Line sensitivity index (LSI) in which three scenarios are considered with line outage and also with different loading conditions, to calculate the power loss. The objective here is the minimization voltage deviation and reduction of power losses in the network to improve the system performance. The methods are applied to the IEEE 14 bus test system.

In modern power distribution networks, Distributed Generation (DG) significantly enhances system performance and operational efficiency. Demand Side Management (DSM) serves as a strategic approach to optimize DG deployment. This study introduces a novel optimization framework that integrates DSM with demand response mechanisms and DG placement. The methodology in this paper identifies optimal DG capacities and locations by evaluating real and reactive power losses alongside voltage profiles. To ensure environmentally sustainable and economically viable operations, the system's daily performance is optimized both with and without demand response integration. A Non-Dominated Sorting Firefly Algorithm (NSFA) is employed for multi-objective optimization, while a fuzzy decision-making model selects the most suitable solution from the Pareto front. The proposed approach is validated using the Practical 32-bus test system, demonstrating its effectiveness through simulation results.

The high rate of the growth of cloud computing has shown concerns of data insecurity, integrity, and transparency in the centralized environment. The current study is a proposal of a cloud architecture that incorporates blockchain to improve secure and transparent data handling by using decentralized verification and cryptography defense. The system uses a combination of the hash function of SHA-256 to verify data integrity, the encryption function of RSA-2048 to ensure secure communication, Practical Byzantine Fault Tolerance (PBFT) to achieve efficiency in the consensus, and authenticated controlled access with the help of smart contracts. Experimental analysis was performed in terms of 10,000 simulated enterprise records in conditions of different workloads. Findings indicate that the presented framework has 100 per cent data integrity detection, 1 per cent lower unauthorized access rates than 8 per cent with traditional cloud systems, and enhances more audit transparency of 98 per cent versus 70 per cent. Even though degradation due to increased transaction latency (290 ms to 360 ms) and slight performance degradation in transaction rate (130 TPS to 115 TPS) were observed under medium load, the security gains and fault tolerance are of much greater importance than the performance overhead. Resilience was tested through the three faulty nodes that had 97% availability in the framework. In general, the paper confirms that blockchain-cloud integration offers a stable yet scalable approach to determining secure and transparent data management with the protection against tampering in the context of distributed computing.

The study of intergroup relations begins with an understanding of the group itself.Ageneral genetic testing of stress genes has been carried out through neuro genetic profiling, where the relational aspects of stress variations between people has been understood in terms of stress predispositions and epigenomics in terms to measure the stress factors in the individual when with groups under stress which are caused through conflicts, thus with basic epicentres indicators that directly or indirectly affect the genes internally through stress factors has been ideated and in this research the supportive categorisation of theory related to group dimensions and conflicts has been analysed.the derivation of standard stress scales seems to help in determining the standard stress level formula to categorise people based on their stress health and variance towards them could help us in understanding their health which was proposed with INGCPT AND SEETHA framework along with neurogenetic profiling of humans that can address the stress of the employee in the organisations. In this study various group settings and related theories from literature has been analysed with support of emotional indicators to adopt the steps involved in invented INGCPT-SEETHA INVENTED NEUROGENETIC PROFILING PROCESS.

In the current scenario customs clearance of goods and procedure services have to undertake both large- scale projects as well as small-scale projects simultaneously. Projects always need to be planned, scheduled, and estimated and finally, those plans need to be executed and the building needs to be constructed at each stage. Customs clearance of goods and procedure projects involves a group of people interacting and being involved in it. When a group of people are working together there are chances for conflict and disputes among the people because all the people won’t think in the same way and thoughts differ between peoples. Every level of the customs clearance of goods and procedure project involves a group of people who needs to be communicated and there should be proper interaction between people to have decision-making. Considering the partnering James Barlow (1998) insisted that knowledge can be improved and shared only through communication. He also stated that communication can solve the complexity and the conflicts involved in the project and commutation flow should be two-way speaker and the listener should involve themselves in the communication properly. When conflict arises when people’s decisions need to be analysed and the reason for the conflicts needs to be identified, which seem to be very vivid tasks. Most of the common reasons for the cause of conflicts are people. Hence people need to be understand analaysed and should be properly handled to resolve the conflicts in the project. If a conflict arises, nowadays people are adopting processeses negotiation, mediation, and arbitration to resolve the conflict but they forget about the people, who seem to be the important reason for conflicts.

AI-Enabled Smart Patient Registration System designed to automate and optimize patient on boarding, clinical triage, and health record management in healthcare facilities. The proposed system integrates secure identity authentication, real-time vitals acquisition, and predictive symptom analytics to intelligently recommend appropriate medical departments and generate digital appointment tokens. Machine learning–based symptom classification models analyze patient-reported symptoms and vital parameters to improve triage accuracy and reduce manual intervention. The system is implemented using modern web technologies to ensure scalability, responsiveness, and cross-platform accessibility, while maintaining centralized electronic health records for longitudinal patient data management. Experimental evaluation conducted in a simulated hospital environment demonstrates a reduction in average patient waiting time by approximately 35–45%, improvement in registration data accuracy by over 98%, and department recommendation accuracy exceeding 90%. User feedback indicates enhanced usability and workflow efficiency for both patients and healthcare staff. The results validate the effectiveness of AI-driven automation in improving operational efficiency, clinical decision support, and patient experience, highlighting the system’s potential for deployment in small- to large-scale healthcare institutions.

Fertilizer recommendation plays a vital role in improving agricultural productivity and ensuring sustainable farming practices. This study applies machine learning models, including Logistic Regression, Decision Tree, Naive Bayes, and Support Vector Machine (SVM), to recommend suitable fertilizers based on key parameters such as soil nutrients (Nitrogen, Phosphorus, Potassium), soil moisture, temperature, and humidity. After data preprocessing and feature selection, model performance is evaluated using accuracy-based metrics to identify the most effective prediction model. Results indicate that machine learning approaches provide more accurate and consistent fertilizer recommendations compared to traditional methods. The system integrates real-time environmental data and provides water level suggestions and fertilizer price estimation to support cost-effective decision-making. This research demonstrates the effectiveness of machine learning in precision agriculture and suggests future enhancements through advanced models and expanded real-time data integration.

Biomedical Named Entity Recognition (BioNER) is vital for extracting structured information from vast amounts of unstructured biomedical text. This study introduces an enhanced BioNER approach by fine-tuning the BioBERT model using adversarial training techniques—specifically, Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD)—to bolster model robustness against input perturbations. By incorporating optimized token alignment strategies, the proposed method significantly improves the identification and classification of biomedical entities across multiple benchmark datasets, including MedMentions, BC5CDR, and i2b2 2010. Comprehensive evaluations using metrics such as Precision, Recall, F1-score, and Entity-Level Accuracy demonstrate that the model consistently surpasses current state-of-the-art systems. This work not only highlights the advantages of adversarial training for domain-specific language models but also sets a new standard for robust and accurate biomedical NER systems.

India is a major silk producer and buyer; hence silk manufacture is vital to its economy. Given India's substantial reliance on agriculture for GDP growth, sericulture - the process of raising silkworms to produce raw silk—is a vital economic activity. The main steps in sericulture are growing plant life for silkworms, spinning silk cocoons, reeling silk filaments, and weaving them into textiles. Silkworm diseases account for 30-40% of production losses. Silkworm illnesses include Grasserie, which is frequent. Silkworm diseases are diagnosed using several medical and laboratory methods. Technology is changing many aspects of our lives, and the agriculture sector has welcomed it. Sericulture has delayed implementing such advancements. While numerous methods have emerged for detecting silkworm eggs and moths, early detection remains difficult. Early disease detection can help farmers prevent disease spread. This research focuses on silk-worm diseases utilizing photo-type and deep-learning models. A device mastery system has been taught to distinguish healthy and unhealthy silkworms using a deep neural network and a CNN, with promising accuracy. TensorFlow was used to create layers and train the algorithm to learn the version. In conclusion, this study uses image categorization and deep learning to detect silkworm diseases. CNN in a deep neural network and TensorFlow have allowed a device to learn a version that can categorize healthy and harmful silkworms.

Emotion AI and Machine Learning, commonly referred to as Affective computing, is an emerging branch of artificial intelligence that enables machines to understand, interpret, and respond to human emotions in a meaningful way. By integrating principles from computer science, psychology, and data analytics, this field aims to create more natural and empathetic interactions between humans and computer systems. One of the primary approaches in emotion recognition involves analyzing facial expressions captured through a camera to identify emotional patterns such as happiness, sadness, anger, surprise, and stress. The proposed application focuses on real-time facial emotion detection using machine learning algorithms that extract facial features, classify emotional states, and display the calculated probability percentages of each detected emotion. By presenting emotion scores in percentage form, the system provides a clear and quantitative understanding of the user’s emotional state. Such technology has significant practical applications in areas including human-computer interaction, customer behavior analysis, online learning engagement, healthcare monitoring, and mental health assessment, thereby contributing to the development of intelligent, responsive, and emotionally aware systems.

The rapid diffusion of voice-enabled and conversational systems has intensified the need for tightly integrated architectures that combine Automatic Speech Recognition (ASR), semantic representation mechanisms, and Large Language Models (LLMs). Although each component has advanced substantially, their structural interdependencies within Automated Voice Query Response Systems (AVQRS) remain underexplored. This study investigates the relationships among ASR quality, semantic embedding fidelity, query clustering coherence, LLM integration quality, and overall system performance. It further evaluates whether LLM model scale (parameter size) and context-window capacity moderate response quality and latency. A random subsample of 300 cases was drawn from a synthetic benchmark dataset (N = 1,000) comprising 20 measured indicators representing latent AVQRS constructs. Statistical procedures included descriptive analysis, Pearson correlation, multiple regression (R²), independent-samples t-tests, one-way ANOVA, and Shapiro–Wilk normality testing across six predefined hypotheses. Results indicated normally distributed performance scores (W = 0.988, p = .882). Construct means were high (e.g., ASR accuracy M = 0.899; semantic embedding quality M = 0.846; overall performance M = 0.864), suggesting strong component-level functionality. However, inter-construct correlations were weak, and regression explained less than 1% of variance in system performance (R² = .005). Neither LLM model size (≤13B, 14–35B, >35B) nor context-window configuration (2,048–8,192 tokens) significantly influenced performance or latency. The findings suggest that isolated component metrics may not capture integrative system dynamics. The study underscores the necessity of theoretically grounded operationalisation and end-to-end, task-based evaluation frameworks, recommending confirmatory structural equation modelling using real-world deployment data.

Forest fires are among the most destructive natural disasters, causing severe environmental damage, biodiversity loss, air pollution, and threats to human life and property. Climate change, rising global temperatures, and prolonged dry seasons have significantly increased the frequency and intensity of forest fires worldwide. Traditional forest monitoring systems rely on manual patrolling and watchtowers, which often result in delayed detection and slow emergency response. This paper presents an IoT-Based Forest Fire Early Warning System designed for real-time environmental monitoring and automatic hazard detection. The system utilizes an ESP32 microcontroller integrated with temperature, humidity, smoke (MQ2), and flame sensors to continuously monitor forest conditions. When environmental parameters exceed predefined safety thresholds, the system triggers immediate alerts through a buzzer and sends notifications via cloud integration. The system also supports GPS-based location tracking to accurately identify fire-prone areas. Experimental testing demonstrates reliable detection of abnormal temperature rise and smoke presence with quick response time. The proposed system provides a cost-effective, scalable, and efficient solution for early forest fire prevention and disaster management.

The rapid growth of digital healthcare services has created a demand for efficient and user-friendly appointment management systems. This paper presents the design and implementation of a Doctor Booking System, a web-based platform developed to streamline the process of scheduling medical appointments between patients and healthcare providers. The system enables patients to register, search for doctors by specialization, view available time slots, and book appointments in real time. Doctors can manage their schedules, update availability, and access patient booking information through a secure interface. The proposed system utilizes a centralized database to maintain patient records, appointment details, and doctor profiles, ensuring data consistency and reliability. Secure authentication and role-based access control mechanisms are incorporated to protect sensitive medical information. The platform reduces administrative workload, minimizes appointment conflicts, decreases patient waiting times, and improves overall healthcare service efficiency.

The use of Virtual Laboratories (VLs) in science education has become increasingly important because they improve understanding, accessibility, and engagement among learners. While earlier research focused mainly on students' achievements and attitudes, there are few studies that examine how students perceive Virtual Laboratories and how this affects their interest in science. The present study aimed to (i) assess students’ level of perception toward VLs, (ii) examine their level of interest in science, (iii) determine differences in perception based on academic stream and achievement, and (iv) explore the relationship between perception and interest. A quantitative descriptive survey design was used for the study. The sample included 100 science students chosen through stratified random sampling from five higher secondary schools of Jammu district of Jammu & Kashmir. Data were collected using a self-made five-point Likert scale that measured perception across five dimensions: perceived usefulness, effort expectancy, motivation and engagement, attitude and preference, and content delivery. A self structured interest scale was also included. Statistical methods such as percentage analysis, t-test, ANOVA, and Pearson’s correlation were applied. The findings showed that most students had high to very high levels of perception about virtual laboratories and interest in science. A strong positive correlation (r = 0.617, p < .01) was found between students’ perception towards virtual laboratories and interest in science. Significant differences in perception were noted across academic streams and levels of interest, but no significant difference was found regarding academic achievement. The study concludes that a positive perception of virtual laboratories greatly improves students’ interest in science, emphasizing the teaching role of integrating virtual lab technology in higher secondary science education

This study examines whether asset quality and profitability differ by ownership among banks operating in Kerala, and whether weaker asset quality is linked to lower profitability. Using a secondary, cross-sectional design for the latest completed financial year, a set of scheduled commercial banks was analysed. Gross non-performing assets, defined as the proportion of impaired advances, and return on assets, defined as net income over average assets, were taken directly from bank disclosures. Sector differences were tested with Welch’s tests, and the asset-quality–profitability link was estimated using Pearson correlation and a simple linear model. All analyses were conducted in EDUSTAT. Private sector banks showed higher profitability than public sector banks, while the gap in gross non-performing assets, though directionally larger for public sector banks, was inconclusive. Across banks, poorer asset quality related strongly and negatively to profitability, consistent with higher credit costs and lost interest income. The narrow scope and unbalanced sample warrant caution, and an annualised quarterly profitability figure for a merged institution may affect comparability. Even so, the evidence supports early risk detection, disciplined provisioning, and timely resolution, and offers a baseline for future multi-year extensions.

This report presents an advanced Smart Irrigation System (S-iS) that leverages state-of-the-art artificial intelligence (AI) and the Internet of Things (IoT) to address global water scarcity in agriculture. The proposed system integrates next-generation multispectral and soil nutrient sensors with a hybrid communication architecture, utilizing NB-IoT/LTE-M for reliable wide-area connectivity and a LoRa-based mesh network for resilient, long-range field coverage. Powered by sustainable solar-energy harvesting, the system employs edge AI processing and fog computing to enable real-time, autonomous irrigation decisions at the node level, significantly reducing latency and bandwidth dependency. Advanced AI methodologies, including reinforcement learning and digital twin simulation, dynamically optimize water delivery based on real-time agro-environmental data, predictive weather modelling, and crop health analytics. Furthermore, the framework incorporates blockchain-verified data integrity and end-to-end encryption to ensure security and trust. Designed for scalability and farmer-centric usability through intuitive mobile and AR interfaces, this solution delivers a substantial reduction in water usage, enhanced crop resilience, and a measurable improvement in the water-energy-carbon nexus, presenting a transformative, sustainable model for precision agriculture in water-stressed regions worldwide.

Online learning has become a core instructional modality in higher education and an important driver of smart and green university initiatives by enabling digitally mediated teaching, institutional efficiency, and reduced reliance on physical infrastructure. Despite increasing investment in educational technologies, the effectiveness of online learning largely depends on educators’ performance and the human capital conditions that support technology-enabled instruction. This study examines how human capital dimensions predict educators’ performance in online learning within higher education institutions. A quantitative survey design was employed, and data were collected from university teaching staff. The data were analyzed using partial least squares structural equation modeling (SEM-PLS). Human capital was operationalized through five dimensions: individual capability, individual motivation, leadership, organizational climate, and workgroup effectiveness, while educators’ online learning performance served as the outcome variable. The results indicate that all five human capital dimensions have positive and statistically significant effects on educators’ online learning performance. Workgroup effectiveness emerged as the strongest predictor (β = 0.367), followed by individual motivation (β = 0.350), individual capability (β = 0.247), leadership (β = 0.206), and organizational climate (β = 0.195). The model explains a substantial proportion of variance in educators’ online learning performance (R² = 0.760). These findings highlight the importance of integrated human capital strategies—including capability development, motivational support, effective leadership, a supportive organizational climate, and collaborative work practices—to enhance online learning quality and support the realization of smart and green university goals.

The Pythagorean theorem is one of the most important mathematical concepts, yet it poses a problem in learning. It is proven that students’ mathematical creative thinking abilities related to this material are still low. There have been quite a few studies on the Pythagorean theorem, but not many have focused on learning development. Therefore, this research aims to optimize students’ mathematical creative thinking abilities through the development of open-ended problem-solving-based differentiated learning. The research design used for development was the Plomp model. The subjects in this study were thirty junior high school students, three mathematics education experts, and one mathematics teacher. The instruments used were an expert validation sheet, a learning response questionnaire, and a mathematical creative thinking ability test. Data was analyzed by combining qualitative and quantitative data analysis. The research findings revealed that open-ended problem-solving-based differentiated instruction was capable of optimizing students’ mathematical creative thinking abilities in learning. This was because differentiated learning facilitated cognitive diversity with various new and creative strategies when solving problems, making learning more meaningful. This research recommended that students be given more opportunities to explore strategies when solving problems.

The current educational administration system still focuses on technical-biographical aspects and has not yet internalized core social values (such as collective responsibility, communication, and mutual trust) throughout the educational process. This gap has the potential to produce graduates who are individually competent but vulnerable to building a holistic safety culture in a stressful work environment. The type of research is qualitative with a field research approach. The data sources in this study are divided into two, namely primary data consisting of lecturers and students and secondary data related to supporting data or documents. The results of the study indicate that: 1). The Configuration of Social Values in Maritime Education Administration is that core social values such as Collective Responsibility, Assertive Communication, and Mutual Trust have begun to be configured into the core of educational administration. 2). The Dynamics of Implementation and Challenges of Administrative Transformation is that in its implementation, this transformation faces complex dynamics and challenges. The main challenge is the clash of paradigms between the flexibility of implementing social values and the characteristics of a rigid and measurable bureaucracy. 3). The Impact of Transformation on the Construction of Cadet Safety Culture is that the impact of the ongoing transformation is beginning to be seen in the construction of cadet safety culture. There is a shift in mindset from reactive compliance to proactive safety awareness, marked by the courage of cadets to "stop work" that is considered unsafe. Social cohesion and effective communication between cadets also strengthened, eroding the oppressive culture of seniority into a relationship of mutually caring partnership.

The heavily technical focus of maritime education has obscured the development of social leadership competencies, which are equally crucial for a seafarer. There is a gap between the internationally certified curriculum and the actual needs on board: the ability to lead multicultural teams, communicate effectively, and manage conflict under pressure. This study employed a qualitative field research approach. The techniques used refer to the criteria of credibility, transferability, dependability, and confirmability. The results of the study indicate that: 1) The administrative configuration indicates that social leadership development planning is still fragmented and not supported by adequate budget allocation. Integration into the curriculum tends to be implicit with minimal portions, while non-academic programs such as LDK lack dedicated funding. Soft skills development is a secondary priority compared to technical training, which accounts for 70% of the training budget. 2) Implementation and field dynamics: program implementation faces the challenge of a hierarchical culture that conflicts with social leadership values. Role disorientation occurs between units, with lecturers focusing on technical aspects while leadership development is considered the responsibility of student affairs units. The heavy technical study load limits cadet participation, so the program only reaches a handful of intrinsically active cadets. 3) Evaluation and reorientation: The existing evaluation system is superficial, measuring only participant satisfaction without measuring long-term impact. There is no formal assessment of the development of social leadership competencies in academic transcripts, thus affecting cadets' motivation for self-development. A complete reorientation towards a holistic assessment system and formal recognition that is commensurate with academic achievement is needed.

Urbanization has significantly influenced the land use and impervious surface growth, thereby altering the hydrology of urban sub-basins or watersheds within urban area. Sub-basins, are highly sensitive to localized land use changes due to their scale and proximity to surface water systems and its influence to change of infiltration and ground water recharge. Land use transformations, particularly those involving conversion of vegetative or permeable land to built-up areas, directly impact infiltration rates, surface runoff, and pollutant transport within these sub-basin. This review paper aims to provide few concepts and approaches for looking to the framework for responsive sub-basin planning by combining insights from 160 scholarly papers across a wide range of geographies, techniques, and temporal scales. The study categorizes present research into policy frameworks, vulnerability assessment, LID (Low Impact Development), hydrological modeling, land use changes, and impervious surface impacts. Apart from identifying methodological gaps and emerging area of discussions in change of urban land use and its effect on urban river, this paper identifying few major parameters which are required to be analyzed for understanding the relationship of urban river and changes of urban land use in a sub basin.

This paper presents the Dynamic Cost-Deviation Greedy Algorithm (DCDGA), a novel algorithmic framework for solving large-scale assignment problems, a class of combinatorial optimization challenges central to theoretical and applied computer science. Unlike the Standard Greedy Algorithm (SGA), which is efficient but often suboptimal, DCDGA integrates a cost-deviation adjustment mechanism, adaptive weighting, and entropy-based prioritization to improve both stability and accuracy. Theoretical analysis demonstrates that DCDGA preserves the asymptotic complexity of the original greedy method (O(n² log n) time, O(n²) space), while offering provable improvements in expected solution cost and robustness under high-variance conditions. Experimental evaluation on synthetic datasets inspired by train scheduling and resource allocation shows that DCDGA achieves, on average, a ~10% reduction in solution cost compared to SGA, with only modest runtime overhead. While SGA occasionally outperforms DCDGA in smaller or low-variance cases, the overall trend indicates that DCDGA is more effective in large-scale, high-complexity problem instances. Comparisons with metaheuristics such as Genetic Algorithms and Simulated Annealing further establish DCDGA's superior scalability for larger datasets. These findings position DCDGA as a practical, scalable optimization technique applicable to diverse domains including logistics, supply chain management, and network design, contributing to advancing research in algorithms and complexity within computer science.

Reductive amination of carbonyl compounds continues to be a fundamental strategy for the synthesis of structurally diverse amines, which are key intermediates in medicinal chemistry and the preparation of complex natural products. In this study, we present an efficient protocol for the reductive amination of aldehydes with aryl amines using a dual catalytic system based on Scandium(III) triflate [Sc(OTf)3] supported on acidic silica. The Lewis acid–mediated activation of the carbonyl group is coupled with Hantzsch-1,4-dihydropyridine (HEH) as a biomimetic organic hydride donor. Under optimized conditions in toluene at room temperature, a wide range of secondary amines were obtained with excellent chemoselectivity and high isolated yields. The protocol proceeds under mild conditions, avoids harsh reagents, and exhibits broad functional-group tolerance, highlighting its potential as a practical and environmentally benign approach for amine synthesis.

A new technique has been developed for the purification of industrial wastewater by synthesizing an innovative Gaunidine based terpolymer composite gaining a considerable attention in the wastewater treatment community. In the present research work the terpolymer resin was synthesized involving 4-Hydroxybenzaldehyde and Gaunidine with formaldehyde and the novel composite was prepared using terpolymer and activated charcoal for the recovery of toxic and heavy metals from aqueous solutions. The structure and properties of the terpolymer and terpolymer activated charcoal composite were observed by various characterization techniques such as elemental analysis, FTIR, UV–Visible, XRD and SEM. Batch separation technique has been used in the ion-exchange process for the removal of heavy selected divalent metal ions like Cu2+, Zn2+, Co2+, Cd2+, Pb2+ by terpolymer and its composite. The investigation was carried out over a range of concentrations, different electrolytes, a wide pH range, and varying flow rates. The selectivity order for metal ion removal by the terpolymer was Zn²⁺ > Cu²⁺ > Co²⁺ > Pb²⁺ > Cd²⁺, whereas for the composite it was Pb²⁺ > Cd²⁺ > Cu²⁺ > Co²⁺ > Zn²⁺. The variation in selectivity order may be attributed to differences in particle size, high porosity, large surface area, and the intrinsic properties of both the material and the metal ions. The ion-exchange behaviour of the terpolymer and its composite was further evaluated. The thermal degradation behavior of the terpolymer and its composite was also examined using thermogravimetric analysis (TGA). The kinetic and thermodynamic parameters were calculated by applying the Freeman–Carroll (FC) and Sharp–Wentworth (SW) methods. The decomposition of the terpolymer followed first-order kinetics, whereas the composite exhibited a higher-order reaction. The calculated activation energy, frequency factor, and entropy change values indicate that both the terpolymer and its composite possess good thermal stability.

Using 2 M HCl as a catalyst, the copolymer was synthesized by condensation of sulphanilic acid (SA), thiourea (T), and formaldehyde (F) in a 1:1:2 monomer ratio. Metal complexes were synthesized by employing this terpolymer as a ligand with two transition metal ions, Cu(II) and Ni(II), in a 2:1 molar ratio. The reaction was run for three hours at 60 °C with effective reflux. UV-visible spectroscopy, NMR, FTIR, SEM, and XRD were used to analyze the resulting metal complexes. The elemental composition of the SATF-I-M copolymeric metal complexes was examined using elemental analysis. The thermal durability of the terpolymer ligand metal complexes was assessed using thermogravimetric analysis (TGA), and the activation energy was calculated using the Freeman–Carroll and Sharp–Wentworth techniques based on TGA data. The RF-501 (PC) S CE (LVD) MODEL PL spectrometer was used to measure the spectra of complexes containing the two transition metal ions in order to study the photoluminescence characteristics of the newly generated copolymeric metal complexes. The main goal of this study is to create new polymeric metal complexes and investigate their photoluminescent properties, while recognizing the important contributions of current researchers in the area.

Present paper deals with the chemical quality of groundwater in the Deccan basaltic aquifer. The Deccan basalt is an igneous rock form by the molten rock material called as magma/lava. Rock is an aggregate of minerals and every mineral has its own fix chemical composition. Chemical composition of basalt reveals SiO2, as major component (48 to 49%), along with the other major chemical components as Al2O3 (13 to 15%), Fe2O3 (12 to 15%), MgO (5 to 7%), CaO (10 to 12%). Minor chemical components comprises Na2O (2 to 3%), K2O (0.1 to 0.5 %), TiO2 (2 to 3.5 %), P2O5 (0.1 to 0.3 %) etc. Results of the hydro-geochemical analysis of groundwater samples from Pewtha area reveals TDS 856 mg/l, Na+ 94 mg/l, K+ 12 mg/l, Ca+ 80 mg/l, Mg+ 77 mg/l, HCO-3 244 mg/l, Cl- 188 mg/l, SO42- 98 mg/l, NO3- 90 mg/l and F- 0.90 mg/l. The careful observation of chemical components and correlation reveals that most of the elements in groundwater are inherited from the chemical composition of the host rock (basalt) and some are due to anthropogenic activities. Thus, present study demonstrate the dominance of chemical components from parent rock, which dissolves in the groundwater and which can vary from place to place depending on rock types.

Different soil types have their own distinct impact on the groundwater levels, seasonal water table fluctuations and yield of the open dug wells. The aim of this study was to evaluate the role of different soil types and their possible implications on groundwater regime in unconfined aquifers. This paper is concerned with the static groundwater levels, water table fluctuations and yield of the open dug wells in different types of soils. The results indicate: (A) Clayey soil represents 12.2 to 13.11 mbgl pre-monsoon SWL, 7.02 to 8.54 mbgl post-monsoon SWL, 4.57 to 5.18 meter water table fluctuation and yield of the dug wells between 36 m3/day to 90 m3/day in pre-monsoon respectively; (B) Fine Calcareous soil represent 7.02 to 16.16 mbgl pre-monsoon SWL, 3.66 to 9.6 mbgl post-monsoon SWL, 1.52 to 6.71 meter water table fluctuation and yield of the dug wells between 36 to 72.9 m3/day in pre-monsoon respectively; (C) Loamy soil and clayey soils (mixed) with moderate erosion (mixed) represent 4 to 18.9 mbgl pre-monsoon SWL, 3 to 12.93 mbgl post-monsoon SWL, 3 to 12.5 meter water table fluctuation and yield of the dug wells between 9 to 97.2 m3/day in pre-monsoon respectively; (D) Loamy soil represent 5.79 to 8.54 mbgl pre-monsoon SWL, 2.15 to 4.27 mbgl post-monsoon SWL, 1.52 to 6.39 meter water table fluctuation and yield of the dug wells between 33.75 to 67.5 m3/day in pre-monsoon respectively. Thus out of four types of soil zones, C- soil zone (loamy + clayey soil) represent deeper groundwater levels, probably due to deeply weathered condition. The water table fluctuation (WTF) also indicates similar results where, C-soil zone has greater WTF values as compared to A, B and D soil zones. On the other hand yield of the open dug wells, tapping unconfined aquifers has higher values in B, C and D soil zones, which indicates negligible control of soil zone on yield of the dug wells. This deviation (higher yield values) may be due to other hydrogeological characteristics of the aquifer and not influenced by soil zones.

Graphene, a two-dimensional allotrope of carbon composed of a single atomic layer arranged in a hexagonal honeycomb lattice, exhibits exceptional electronic and optical properties arising from its linear energy dispersion and massless Dirac fermions. Since its experimental isolation, graphene has attracted intense research interest due to its broadband optical absorption, ultrafast carrier dynamics, high optical conductivity, and strong nonlinear optical response. The rapid and multidisciplinary growth of literature on the optical properties of graphene has resulted in a large, heterogeneous knowledge base, making it increasingly difficult to systematically identify dominant research themes, emerging trends, and knowledge structures. To address this gap, the present study performs a comprehensive keyword-based scientometric analysis of research on the optical properties of graphene. A total of 1,196 documents indexed in the Web of Science (WoS) database from January 2004 to March 2025 were analysed employing three different approaches of keyword analyses viz. keyword co-occurrence analysis, cluster analysis of thematic map and factorial analysis. Visualization and mapping were carried out using VOSviewer, while thematic evolution was examined using the Bibliometrix R package. The analysis reveals the conceptual organization of the field, highlights core and emerging research clusters, and traces the intellectual evolution of graphene optics research. The findings underscore the central role of graphene's optical properties in enabling applications such as photodetectors, optical modulators, saturable absorbers for ultrafast lasers, transparent electrodes, plasmonic and optoelectronic devices, and broadband optical sensors. This study provides a structured overview of the research landscape and offers valuable insights for researchers and academicians aiming to guide future developments in graphene-based optical and photonic technologies. Furthermore, the dominant keyword clusters are mapped to real-world graphene-enabled products, illustrating the translational progression of fundamental research toward commercial technologies.

This work analyzes a glass sample with the chemical formula (65-x)B2O3-15TeO2-15Bi2O3-5La2O3-xAg2O, where x equals 0, 0.5, 1, 1.5, and 2 mol%. The density of the glass samples increased from 3.15 to 3.2 g/cm³, and the molar volume also increased from 49.326 to 49.56 cm3/mole, with increasing molar concentration of Ag2O. The linear attenuation coefficients were calculated for gamma photons over the energy range 015MeV – 15 MeV. With these linear attenuation coefficient values, further attenuation parameters, i.e., half value layer (HVL), mass attenuation coefficient (MAC), and effective atomic number (Zeff), were then calculated. The measured values of these attenuation parameters were also compared with the theoretical values obtained through the online software Py-MLBUF. The findings show that the linear attenuation coefficient (LAC) increases with increasing sample density and increasing proportions of Ag2O. The prepared glasses indicate a stronger shielding ability among the materials investigated than the usual glass and concrete samples do. This study highlights the potential of these glass compositions as effective materials for gamma radiation shielding.

In this study, Lithium Zinc telluroborate glasses incorporated with Lanthanum nanoparticles (LZTBLn) were prepared using the conventional melt-quenching method. The density and molar volume of the glasses were calculated. XRD analysis confirmed that all samples are amorphous in nature. The radiation shielding ability of the glasses was evaluated using computational simulation technique. Shielding parameters such as the Linear attenuation coefficient (LAC), Mass attenuation coefficient (MAC), Effective atomic number (Zeff), Half-value layer (HVL), Tenth-value layer (TVL), and Mean free path (MFP) were determined over a wide range of photon energies (0.015-15MeV). The results show that increasing the La2O3 nanoparticles content improves the radiation shielding performance than other traditional shielding materials. These findings suggest that La2O3 doped lithium zinc telluroborate glasses are promising, lightweight, and non-toxic materials for radiation shielding applications.

A new set of glass samples consisting of 30Na2O-5PbO-5PbF2-10BaO-(60-x)B2O3-x Pr2O3, with 0 ≤ x ≤ 0.5, was synthesised using the melt quench technique. The radiation attenuation characteristics of the produced glasses were analyzed using Phy-X/PSD software. The values of μ increased from 0.015 MeV, where the MAC values are 37.008 cm2/g and 37.522 cm2/g for BaO at 10 mol% with Pr2O3 at 0% and 0.5%, respectively. The HVL value for BaO-Pr-0.0 glass was measured at 3.702 cm when subjected to 1.5 MeV, and BaOPr-0.5 glass demonstrated HVL values of 4.069 cm at the same photon energy level. The MFP values measured reflect HVL characteristics, indicating that BPr-5.0 glass provides superior values across different photon energies when assessed against the other prepared glasses. Among the tested materials, prepared glasses reveal enhanced shielding effectiveness when juxtaposed with commonly utilized glass and concrete samples. The findings of this study highlight the feasibility of these glass types as practical solutions for shielding against gamma radiation.

Rare earth infused Borobismuth glasses were synthesized using traditional melt quenching experimental method to study their structural and luminescence properties. Amorphous nature of the samples was confirmed by X-ray diffraction and Raman analysis revealed progressive recognization of borate structural units resulting in modified network connectivity. The observed variations in density, optical polarizability, and related physical and optical parameters confirm the stabilization of the glass network compactness. Photoluminescence (PL) investigations showed broad ultraviolet visible emission with intensity modulation controlled by heavy metaloxide and rare earth concentration. An improvement in emission efficiency at lower concentrations is associated with enhanced structural asymmetry and stronger local field effects, whereas luminescence quenching at higher concentration is linked to increase the non radiative energy transfer. These findings indicate that the prepared rare earth infused borobismuth glasses may be used in solid state lighting and advanced photonic applications.

Natural rubber (RSS1 grade) reinforced with silica derived from bamboo leaves ash (BLA) was developed as a sustainable biocomposite material. The biosilica was obtained through controlled calcination of bamboo leaves followed by purification to yield predominantly amorphous SiO₂. Composites containing varying filler loadings were prepared through two roll mill mixing and subsequent vulcanization. Structural interactions between the rubber matrix and bio-silica were investigated using Fourier Transform Infrared (FTIR) spectroscopy. FTIR spectra revealed characteristic absorption bands of cis-1,4-polyisoprene along with Si–O–Si stretching vibrations, confirming the successful incorporation of silica without altering the fundamental structure of the rubber backbone. Thermogravimetric analysis (TGA) demonstrated improved thermal stability and increased char residue with increasing biosilica content, attributed to the barrier effect and restricted polymer chain mobility. The results indicate that bamboo leaf–derived silica can act as an effective eco-friendly reinforcing filler for RSS1 rubber, enhancing thermal resistance while supporting sustainable material development.

Nutrient depletion caused by heavy rainfall and excessive irrigation significantly affects soil fertility and plant productivity. The present study investigated the impact of nutrient loss induced by soil washing on the growth and biochemical performance of Spinacia oleracea (spinach plant) and evaluated the potential of magnesium ferrite nanoparticles to mitigate nutrient deficiency stress. Normal soil (C1) and washed soil (C2) were used as control groups, while washed soil supplemented with magnesium ferrite nanoparticles at 10, 20, 40, and 80 mg per 100 g soil constituted the treatment groups (E1–E4). Plant growth parameters, including germination rate, vigour index, and fresh biomass, along with biochemical parameters such as protein, reducing sugar, chlorophyll, and proline contents, were assessed at 15, 30, and 60 days after germination. Plants grown in washed soil exhibited reduced biomass, chlorophyll, protein, and reducing sugar levels, accompanied by elevated proline accumulation, indicating nutrient-deficiency induced stress. Supplementation with magnesium ferrite nanoparticles significantly improved growth and biochemical parameters in a concentration-dependent manner and reduced proline accumulation compared to washed-soil controls. The findings demonstrate that magnesium ferrite nanoparticles can partially restore plant growth and metabolic performance under nutrient depleted soil conditions, highlighting their potential as a nano-amendment for improving crop productivity in leached soils.

Surface modification using metal oxide materials enhances substrate properties like biocompatibility, corrosion resistance, and antibacterial activity, often applied to nanoparticles or implants such as titanium alloys. Physical methods include plasma spray for dense coatings like hydroxyapatite or Al₂O₃ on titanium, and physical vapor deposition (PVD) for uniform TiN films improving wear resistance. Chemical approaches encompass sol-gel for uniform oxide films like TiO₂, micro-arc oxidation for ceramic layers, and chemical vapor deposition (CVD) for controlled TiO₂ nanowires. Bio-inspired methods use catechol-based dispersants for strong adsorption on oxide nanoparticles, enabling electrophoretic deposition (EPD). Titanium dioxide (TiO₂), iron oxide, alumina (Al₂O₃), zirconia (ZrO₂), and silver oxide (Ag₂O) dominate, often doped with Ag, Sr, or Zn for added bioactivity. Silane coupling agents functionalize metal oxide nanoparticles to reduce agglomeration and improve dispersion in composites. Recent advances include block copolymer-templated metal oxide nanopillars on polymers for tailored surfaces. This article covers the study of functionalizing Fe₃O₄ and ZnO nanoparticles (NPs) for biocompatibility involves coating their surfaces with biocompatible materials to reduce toxicity, prevent agglomeration, and enhance stability in biological environments. Polyaniline (PANI) and aryldiazonium salts (ArN2+) are tried as coatings which minimize immune recognition and cytotoxicity up to considerable level. This surface functionalization improves biocompatibility as well efficacy of metal oxide sensors.

A new chelating ion-exchange copolymer was prepared using the polycondensation of 2,4-dihydroxybenzoic acid, acrylamide and formaldehyde in a 1: 1: 2 proportion. The copolymer produced was widely characterized by UV-Visible spectroscopy, Fourier transform infrared spectroscopy (FT-IR), proton nuclear magnetic resonance (1H NMR), carbon-13 nuclear magnetic resonance (¹³C NMR), scanning electron microscopy (SEM), X-ray (XRD), elemental analysis, and gel permeation chromatography (GPC) of the molecular weight. The identification of the successful formation of the copolymer and the presence of functional groups that were able to coordinate with metal ions was proven by the spectral and analytical findings. The ion-exchange ability of the synthesized copolymer was tested against the removal of Fe³⁺, Cu²⁺, Cd²⁺, Zn²⁺, Ni²⁺, and Pb²⁺ ions in aqueous solutions using the batch equilibrium technique. The influence of the concentration of the electrolytes and the pH and the contact time on the adsorption of metal ions was systematically studied. The findings showed that the copolymer had a high adsorption capacity and selectivity to the toxic metal ions especially Pb²⁺ and Cd²⁺, because of the presence of hydroxyl and amino functional group, which formed stable chelate complexes. The resulting copolymer was found to have a high potential of being an effective and reusable ion-exchange material in the removal of dangerous metal ions in the polluted water systems.

Question answering models are designed to automatically generate accurate and contextually relevant answers from a dataset. These models leverage advanced techniques in natural language processing, such as transformer architectures like BERT and RoBERTa, to understand and extract the precise information needed to formulate responses. This work presents a comparative evaluation of BERT and RoBERTa models, focusing on key performance metrices such as Exact Match (EM), BiLingual Evaluation Understudy (BLEU), and F1 Score. Start word score and end word score determine the precise boundaries of the answer within a passage. These scores reflect the ability of the model and its accuracy is crucial for high EM and F1 Scores. RoBERTa's advanced architecture and fine-tuning processes enable it to more accurately identify these positions, resulting in more precise and contextually relevant answers. This study highlights RoBERTa's superior performance, with an EM of 75%, a BLEU of 80%, and an F1 of 87%, outperforming BERT, which achieved 70%, 75%, and 82% in the respective metrics. The findings of this study establish RoBERTa as the preferred model for question answering tasks, particularly in applications requiring high precision and exact answer identification. This research emphasizes the importance of start and end word selection in driving model performance and suggests areas for further refinement in question answering systems.

Every year, thousands of people die because they don't get medical help quickly enough after a road accident. Road accidents still rank among the top causes of death worldwide, and honestly, the time it takes for someone to alert emergency services can mean the difference between life and death. That's where IoT-based accident detection steps in—it's a smart, automated way to make sure help gets there fast. Here's how it works. The system uses a mix of hardware and communication tools. Accelerometers and vibration sensors keep an eye out for sudden impacts, harsh tilts, or weird movements—things you'd expect during a crash. If the readings go past a certain threshold, the system knows there's been an accident. Right after that, GPS kicks in and pinpoints the exact location. This way, rescue teams know exactly where to go, wasting no time. For sending out alerts, the system uses a GSM module or another communication device. It instantly notifies emergency contacts, hospitals, or control centers and sends them the location data. To avoid false alarms, a microcontroller runs the show, checking everything with logic and validation before calling for help. The paper breaks down the whole system—what it's made of, how it works, and where you can use it. This IoT solution brings together real-time monitoring, automatic alerts, and spot-on location tracking to make roads safer. It speeds up emergency response and helps rescue teams coordinate better, raising the chances that accident victims get the right care in time.

In recent years, effective personal financial management has become increasingly important due to the rise in daily expenses and the need for better budgeting practices. Many individuals face challenges in tracking their income and expenditures, which often results in poor financial planning and unnecessary spending. This paper presents the design and development of a cloud-based financial tracking application that assists users in managing their financial activities in a simple and efficient manner.

The proposed system is developed using Flutter for the frontend and Firebase for cloud-based data storage and real-time synchronization. The application enables users to record income and expenses, categorize transactions, and monitor their financial status through graphical representations such as charts and summaries. The use of cloud technology ensures secure data storage, easy accessibility, and real-time updates across devices.

The system is designed with a user-friendly interface to ensure ease of use for individuals with different levels of technical knowledge. By providing clear insights into spending patterns and financial behavior, the application helps users make informed decisions and maintain better control over their finances. The proposed solution demonstrates how mobile and cloud technologies can be effectively utilized to improve personal financial management.

The current investigation seeks to explore the impact of perceived authenticity of brands on the motivations for making sustainable purchases from Gen Z consumers utilizing the Meesho social commerce website. The research will examine how both authentic communication from brands and lifestyle orientated marketing strategies lead to environmentally sustaining purchasing behaviour. The data for the study was collected through the administration of a cross-sectional survey using a convenience sample, consisting of 186 completed surveys submitted by Gen Z individuals who utilized the Meesho social commerce platform. The results of the analysis were based primarily on, descriptive statistics, the results of reliability testing (Cronbach's alpha), and correlation and linear regression. Therefore, it can be concluded that while brand association between lifestyle and sustainability behaviour is present. While brand cues increase exposure to and facilitate an average level of behaviour change, they do not have a direct impact on sustainable consumption because of many influences in the external environment. Therefore, social commerce should promote authenticity in sustainability messaging, transparency, and community-driven branding in order to build consumer confidence and encourage responsible purchasing decision-making. In summary, perceived authenticity serves as a behavioural catalyst for - but not the primary driver of - sustainable behaviour within Gen Z.

Proper prediction of sugarcane distribution rate based on the stem border features is a difficult task because of the irregular forms and morphology, changes in the environmental conditions, and noise during imaging procedures. The proposed paper is a hybrid machine learning architecture that incorporates Capsule and Graph Neural Networks (GNN), Gaussian Process Regression (GPR) and Extreme Learning Machine (ELM) to simulate structural, spatial, and probabilistic sugarcane stem image modeling. The system was trained on one lakh high-resolution border-segmented images and cross-validation done on a large-scale dataset of these images. The fusion strategy that is suggested will have hierarchical texture coding, graph-based boundary representation, and nonlinear regression refinement to increase predecessence and generalization. Compared to traditional deep learning and regression models, experimental evidence proves a higher degree of predictive stability and accuracy of 99.487. The system is a scalable solution to intelligent agricultural analytics, to which it provides automated grading, yield optimization, and precision farming applications.

Chronic Kidney Disease (CKD) is a chronic type of disease with gradually impaired kidney functioning as its characteristic feature, whereas its global problem has become a critical issue of public health. The World health Organization (WHO) refers CKD to be one of the leading causes of morbidity and mortality globally and in low- and middle-income countries mostly missed at an early stage. This disease is usually asymptotic at the initial stages hence requiring effective and precise predicted procedures that identify CKD before graduating to end-stage renal failure.

Machine Learning (ML) has become a very effective instrument in the field of medicine, especially in terms of disease prediction, in recent years because of its ability to practice sophisticated patterns in substantial volumes of data. ML algorithms Decision Trees, Support Vector Machines, Random Forest, Deep Learning architecture have been used to create predictive models that have potential to help identify personalities that are at risk of having CKD. Moreover, stratification methods, including demographic-, diseases- and lifestyle-related stratification, help improve the specificity and significance of such models because they allow them to be used to assess risks to the individual.

Although ML has potential in the prediction of CKD, issues such as the standardization of data as well as management of imbalanced data, interpretability, and the incorporation of clinical validation need to be addressed. In addition, a possible integration of the new categorization models with the latest ML methods provides a way to new, more stable, more transparent, and clinically acceptable models.

The given systematic review is aimed at a broad analysis of the available literature on the prediction of CKD and refers to the studies that integrate an approach to categorization and the novel machine learning strategies. The review compares the methodologies, performance measures, characteristics of the data and clinical implications to give a summative view which can be used in future research and practical implications.

This paper introduces the α-Laplace Elzaki transform, a significant generalization of the nabla version on time scales. This transform rigorously established the fundamental properties, including the existence theorem, linearity, and the convolution theorem. Additionally, we derive the key derivative characteristics that underpin its functionality. The α-Laplace Elzaki transform combines techniques for discrete, continuous, and hybrid systems and is a powerful approach for solving partial dynamic equations over time scales.

Cyber-Physical Systems (CPS) like smart grids, industrial automation units, and modern manufacturing plants are associated with the need to make decisions in real-time and ensure high security. Machine Learning (ML) has strong detection and prediction functions, yet the conventional inference with the help of a CPU or cloud is too slow and can be compromised. At the same time, the integrity of CPS is severely endangered by compromised ML models or bitstreams of the FPGA. In this paper, the author suggests a FPGA-Accelerated Machine Learning Framework with a Blockchain-Based Integrity Validation as a means of securing CPS environments. The method employs FPGAs to detect anomalies and predictive models at the edge and attains ultra-low latency. A blockchain layer provides tamper-proof model provenance, validation of bitstream and safe event logging. The integrated design is both very reliable, very fast, and decentralizes trust, and is resistant to model pollution and device impersonation. The experimental evidence of using representative ML models illustrates a substantial decrease in the time of inference, latency, and security robustness.

Managing medical images securely and efficiently is a growing challenge, especially as modern imaging technologies produce color and high-bit-depth files that are both large in size and highly sensitive. To address this, a two-stage framework is proposed that combines precision compression with resilient encryption. The beginning stage is Key-Pixel Predictive Modeling for Lossless Compression (KPPMLC), reduces storage requirements by selecting a small number of mandatory pixels and by applying a context-aware neural network to accurately reconstruct the rest. KPPMLC is built completely using deterministic and reversible operations, that approach enables a perfect image reconstruction, with support for both grayscale and RGB formats up to 16 bits per channel. The next stage which is named as Hybrid Quantum-Classical Magic Square Generator (HQCMSG), builds 256×256 cryptographic magic square matrices using a combination of classical backtracking and quantum optimization based on QUBO. A coordinating mechanism keeps monitoring both paths and selects the first valid result, which helps to maximize randomness and to improve the encryption strength. Together, these modules provide a compact, lossless, and future-proof system for the secure handling and transmission of medical images. Evaluation on a CE-T1 MRI dataset (3,064 images) shows that QCMSE achieves bit-exact reconstruction (MSE = 0) for grayscale and RGB images up to 16-bit depth. It reduces average encryption and decryption times to 3962.6 mS and 4,026.6 mS, outperforming existing methods by up to 25.3% and 23.9%, respectively. Strong security is validated by an average NPCR of 99.83%, UACI of 34.15%, and an overall security score of 98.38%, confirming robustness against statistical and differential attacks.

The rapid adoption of Internet of Things (IoT) technologies in smart cities, healthcare, and industrial automation has increased the vulnerability of networked systems to cyber threats. Due to distributed deployment and limited computational resources, IoT environments are susceptible to attacks such as distributed denial-of-service, botnet propagation, unauthorized access, and previously unseen anomalies. Centralized intrusion detection systems are often inadequate because of high latency, scalability challenges, and privacy risks associated with transmitting sensitive traffic data to cloud servers. This paper proposes a federated edge- intelligence framework for real-time cyber threat detection using explainable deep learning. Deep learning–based anomaly detection models are deployed at edge nodes to enable low-latency detection, while federated learning supports collaborative model training without sharing raw data. Explainable artificial intelligence techniques enhance transparency by identifying influential traffic features. Experimental evaluation demonstrates improved detection accuracy, reduced response time, and enhanced interpretability, making the framework suitable for securing next- generation IoT environments.

Detecting copy-move forgery has become an inevitable and crucial task due to the advancements in highly sophisticated image processing tools. The conventional CMFD approaches confront the detection of tampered images among the cross-domain and domain-specific tampered images regardless of the domain knowledge source. Thus, this work proposes two contributions to overcome these constraints: the generalized CMFD and domain-specific CMFD by utilizing the dual feature representation jointly extracted from the handcrafted and deep features. The proposed CMFD incorporates the dual feature extraction, classification, and localization phases. In the generalized CMFD, dual feature representation involves the extraction of hybrid handcrafted features, and EfficientNetV2B3-assisted deep features leverage the accurate detection of tampered images using the Vision Transformer (ViT). Subsequently, it localizes the forged region in the tampered image based on the inter and intra-block mapping in the pixel-level clusters. In the domain-specific CMFD, dual feature representation with VGG16-assisted deep feature extraction additionally employs domain-invariant representation learning for the superpixel-level analysis. It relies on the enhanced Swin Transformer with the efficient local self-attention using shifted windows for capturing global context during tampered image detection. Finally, the domain-specific localization utilizes the domain-invariant feature space with the correlation analysis in addition to the inter and intra-block mapping to detect the forged region during postprocessing precisely. Experimental results demonstrate the notable performance of DFRFD by evaluating the different benchmark image datasets for the generalized CMFD algorithm and domain-specific CMFD algorithm. As a result, the generalized CMFD algorithm yields 97% accuracy while testing on the CoMoFoD dataset.

Early and accurate detection of brain tumours is crucial for improving patient survival and treatment planning. This study presents a framework based on deep learning for automated brain tumour detection and classification using magnetic resonance imaging (MRI) scans. A convolutional neural network (CNN) architecture was developed and trained on a curated dataset of labeled brain MRI images. This model can distinguish between tumour and non- tumour cases, as well as different tumour categories. The proposed model uses preprocessing techniques like image normalization, resizing, and augmentation to improve generalization and reduce overfitting. The network architecture includes multiple convolutional layers with ReLU activation, max-pooling operations, batch normalization, and fully connected layers, followed by a softmax classifier. We evaluated model performance using standard metrics such as accuracy, precision, recall, F1-score, and loss analysis. Experimental results show high classification performance, with strong accuracy and balanced precision-recall characteristics across validation and test datasets. Training and validation curves indicate stable convergence with minimal overfitting. A comparative analysis with baseline machine learning approaches confirms that the proposed deep learning framework is superior in feature extraction and classification robustness. The system has potential for real-time clinical decision support, helping radiologists with early tumour identification and reducing their diagnostic workload. This study highlights the effectiveness of convolutional neural networks in medical image analysis and supports their use in automated brain tumour diagnosis. Future work will focus on multi-class tumour grading and integration with transfer learning techniques. learning models, and validation on larger multi-institutional datasets to improve clinical applicability and scalability.

Efficient irrigation management plays a critical role in sustainable agriculture and water conservation. This paper presents the design and implementation of a cloud-enabled smart irrigation system using the ESP32 microcontroller and the Blynk IoT platform. The system monitors soil moisture, temperature, and humidity in real time using an analog soil moisture sensor and a DHT22 environmental sensor. Based on threshold-based decision logic, a relay-controlled water pump is automatically activated when the soil becomes dry and deactivated when sufficient moisture is detected. The system integrates cloud connectivity through the Blynk IoT platform, enabling remote monitoring and control via a mobile dashboard. It supports both automatic irrigation mode and manual override mode. A 16×2 LCD provides local real-time system feedback. The proposed solution demonstrates a low-cost, scalable, and hybrid edge-cloud IoT architecture suitable for smart agriculture and small-scale irrigation systems.

Aircraft maintenance safety is critically influenced by human and organizational factors, yet maintenance errors continue to represent a significant risk within aviation operations. This study investigates the influence of human performance factors, organizational conditions, and operational pressures on aircraft maintenance errors using an integrated analytical approach. The research adopts a mixed-methods design combining quantitative survey data with qualitative insights from maintenance professionals. Data were collected from 120 aircraft maintenance personnel, including technicians, supervisors, and safety managers, through structured questionnaires, interviews, and analysis of maintenance reports. Statistical analysis was conducted using descriptive statistics, correlation analysis, multiple regression, and chi-square testing to examine relationships between key variables such as human factors awareness, training adequacy, communication effectiveness, workload and fatigue, and organizational safety culture. The findings indicate that workload and fatigue are the strongest predictors of maintenance errors (β = 0.35, p < 0.001), while human factors awareness (β = −0.31, p < 0.001) and organizational safety culture (β = −0.28, p = 0.001) significantly reduce error occurrence. Correlation results further demonstrate that maintenance errors are positively associated with workload and fatigue (r = 0.49) and negatively associated with safety culture (r = −0.57). Additionally, chi-square analysis reveals a significant relationship between training levels and maintenance error reporting (χ² = 14.27, p = 0.001). The study concludes that aircraft maintenance errors arise from the interaction of human, organizational, and operational factors rather than isolated individual failures. Strengthening human factors training, improving communication procedures, implementing fatigue risk management systems, and promoting a proactive safety culture are essential for enhancing maintenance safety. The research contributes to aviation safety literature by providing empirical evidence supporting integrated human–system approaches to maintenance error prevention.

Neural Architecture Search (NAS) has demonstrated remarkable success in automatically designing high-performance deep neural networks. Most NAS algorithms are computationally expensive and they are typically optimized for a single domain, limiting their applicability in real-world scenarios where domain shifts frequently occur. In this paper, we propose MetaDANAS, a novel framework that integrates meta-learning and domain adaptation into neural architecture search to enable efficient architecture discovery across heterogeneous domains. The proposed method learns transferable architectural priors using episodic meta-training over multiple source domains while incorporating a domain adaptation module to align feature distributions between source and target datasets. To enable efficient architecture exploration through weight sharing, a differentiable supernet search space is employed. MetaDANAS significantly reduces architecture search time while achieving superior cross-domain accuracy compared to conventional NAS methods is demonstrated with experimental results across multiple cross-domain image classification tasks. The proposed approach provides an efficient solution for scalable neural architecture design in multi-domain environments.

The psychological health of the students is a very high influencer of emotional stability, behavior and academic performance. The crises should be avoided by detecting the anomalies early. The existing methods face the problem of noisy high-dimensional campus data and fail to pick subtle signs of distress in normal variability. Another model that can help solve these issues is the Temporal Sensitive Network (TSN), a new behavioral time series analysis model that will be used to identify psychological distress. TSN uses a two‑phase pipeline. The initial step involves Jenks natural breaks which are applied to the features to give the discretization of the features and then Apriori is used to mine rules that correlate with the health indicators. This step derives discriminative signals on consumption, internet and activity logs. The second phase uses an attention enhanced gated module that combines long-term habits with short-term variations with more preference to anomalies through soft-max-weighted representations. The outcome is a contextual anomaly detector that is dynamic. Experiments on the Student Life data demonstrate that TSN performs better than the usual baseline (RF, SVM, LSTM, ST -GCN) with the accuracy, precision, recall, and F1 score of 78.4, 77.6, 78.0 and 77.8 respectively. The research of Ablation validates the contributions of every constituent, and interpretability visualizations clarify the decision-making of the model. As shown in this work, a privacy preserving, scalable methodology can be used to facilitate proactive campus support systems.

In casting components, Non-destructive testing (NDT) techniques are widely used to detect and characterize internal flaws. Among various inspection methods, ultrasonic testing (UT) offers high sensitivity to internal discontinuities and provides a safer alternative to radiographic testing. This research investigates the influence of key parameters such as surface preparation, probe frequency, signal interpretation, and material density on flaw detection accuracy. Advanced UT approaches were employed to generate flaw mapping across different materials, where multi-probe scanning demonstrated improved accuracy, efficiency, and suitability for complex geometries. Synthetic Aperture Focusing Technique (SAFT) images were reconstructed from A-scan ultrasonic data to enhance defect visualization. SAFT enables post-processing beam focusing at each image point, leading to improved spatial resolution and clearer identification of internal defects. The results confirm that optimized UT parameters combined with SAFT imaging significantly improve reliability, defect localization, and overall quality assessment of casting components.

In recent trends wood polymer composites (WPCs) have gained significant attention due to their sustainability and balanced mechanical performance, where as poor interfacial adhesion, bonding between hydrophilic wood fibers and hydrophobic polymer matrices often limits their properties strength. The coupling agent maleic anhydride grafted polyethylene (MAPE) is used as a compatibilizer to improve interfacial bonding in WPCs. This study examines the effect of MAPE on the physical, mechanical properties of WPCs. Even less percent of MAPE incorporation significantly enhances tensile strength, flexural strength, and impact resistance due to improved stress transfer between the wood fibers and polymer matrix. Scanning electron microscopy reveals reduced fiber pullout and better fiber dispersion in MAPE modified composites. MAPE reduces water absorption by enhancing interfacial compatibility. Hence, the use of MAPE as a coupling agent effectively enhances the properties and durability of WPCs, making them more suitable for structural and various practical applications.

This paper examines the impact of chemical reaction, radiative heat transfer, and mass transfer on unsteady magnetohydrodynamic (MHD) Casson fluid flow including heat generation and viscous dissipation passing a vertical porous plate. The fluid is taken to be incompressible and electrically conducting, thermal radiation being modelled by the Rosseland approximation. The equations of state that are governing, nonlinear partial differential equations of momentum, energy, and concentration, are postulated under the Boussinesq approximation and scaled into dimensionless equations using appropriate similarity transformations. The ensuing coupled set of equations is numerically solved by the Crank Nicolson finite difference method. The influence of the different physical parameters on velocity, temperature as well as concentration profile is analyzed and discussed with the help of graphical and tabular analysis of the physical parameters which include Casson parameter (β) and thermal Grashof number (Gr), solutal Grashof number (Gc), radiation parameter (R), Prandtl number (Pr), Schmidt number (Sc), chemical reaction parameter (Kc), heat generation parameter (Q) and Eckert number (Ec). Moreover, the skin friction coefficient, Nusselt number and Sherwood number have been calculated and tabulated. The findings also indicate that radiation and chemical reaction have significant effects on thermal and concentration boundary layers and heat generation and viscous dissipation improves the velocity and temperature fields. The research is helpful in applications to industrial cooling, polymer processing, geothermal energy mining and chemical engineering process using non-Newtonian Casson fluids.

The cardiovascular diseases continue to be among the top causes of death in the world, which has necessitated a high demand of constant and efficient health monitoring systems. The proposed research designates a smart wearable e-health monitoring system based on IoT as a device that provides real time readings of physiological parameters and immediate detection of possible heart abnormalities in order to provide heart patients with proper treatment. The system is proposed to combine wearable sensors and Internet of Things (IoT) to gather and exchange patient health information, such as heart rate, blood pressure, electrocardiogram (ECG), body temperature, and oxygen saturation levels. The gathered data is sent to a cloud-based server in which machine learning algorithms are used to define the health condition and predict risk. The analysis of patient health data using four machine learning algorithms was used to detect abnormal cardiovascular patterns among the patients, namely, the support of the Support Vector Machine (SVM), the Random Forest, the K-Nearest Neighbour (KNN), and the Logistic Regression. The experimental test showed that the given system is accurately able to identify health risks associated with the heart. The algorithm model that had the highest performance was the Random Forest algorithm with an accuracy of 96.2, precision of 95.4 and a recall of 94.8 and the SVM with an accuracy of 93.4. The system also produced health alerts with an average response time of 2-3 seconds, thus allowing quick intervention by the medical personnel. The findings suggest that the suggested IoT-based monitoring system can greatly enhance the provision of remote health care, early diagnosis of heart diseases, and the safety of patients by providing them with constant health monitoring.

Psychological stress significantly affects cognitive performance, emotional stability, and overall well-being among students and professionals. Traditional stress detection methods rely on self-reported surveys or physiological sensors, which are often intrusive, costly, and unsuitable for continuous monitoring. This paper proposes a real-time, non-invasive Artificial Intelligence-based framework for automatic emotion recognition and stress level prediction using facial expressions. The system integrates image preprocessing, face detection, and a deep Convolutional Neural Network (CNN) for multi-class emotion classification. Based on detected emotional states, stress levels are inferred using an emotion-to-stress mapping model. Experimental evaluation on benchmark facial expression datasets demonstrates an emotion classification accuracy of 92% and stress prediction accuracy of 90%. The proposed approach offers a scalable and cost-effective solution for stress monitoring in education, workplace, and healthcare environments

This precision in detecting cardiac risk at an early phase is still one of the most important requirements in the clinical decision support. The present paper suggests the development of a comprehensive analytical structure of prediction of heart disease with the help of structured clinical features on the UCI Cleveland dataset. During data preprocessing, feature transformation, feature normalization, and feature encoding as well as dimensionality reduction of features is performed to apply machine learning and deep learning. The comparison of different machine learning classifiers is conducted, and a hybrid deep learning system of CNN and LSTM is created to represent the local dependencies and longitudinal relationships that are inherent to the features of the patient. A comparative evaluation using the traditional machine learning algorithms and deep learning models validates enhanced predictive stability with enhanced generalization results of the hybrid model. It highlights the suitability of the integrated deep learning arrangements in terms of medical risk forecasting using clinically heterogeneous indicators.

Farmers must innovate to adapt to ongoing changes; however, many are becoming psychologically powerless, and farming is increasingly viewed as an unattractive profession among younger generations. This study examined the mediating role of work engagement between psychological empowerment and innovative work behavior among farmers, as well as differences in these variables between younger and older farmers. Data were collected from 403 participants using validated scales for the three variables. Analyses were conducted using Partial Least Squares Structural Equation Modeling (PLS-SEM) with SmartPLS 4.1.12 and the Mann–Whitney U test. The results indicate that psychological empowerment significantly influences both work engagement and innovative work behavior. Work engagement, however, does not significantly influence innovative work behavior and therefore does not function as a mediator. The predominantly top-down pattern of innovation dissemination from extension agents to farmers appears to be a key reason for the lack of a mediating effect. Although innovation is externally introduced, farmers’ engagement emerges as a consequence of implementing these innovations rather than as a driver of innovative behavior. These findings suggest that top-down and bottom-up approaches to promoting innovative work behavior have different impacts on work engagement. Prior studies conducted in bottom-up contexts have shown that work engagement mediates psychological empowerment and innovative work behavior; workers who already value their work tend to show greater innovation. In contrast, in this study, innovation was provided by extension agents, making work engagement an outcome rather than a mediator. The study also found that younger farmers exhibit higher levels of psychological empowerment, work engagement, and innovative work behavior compared to older farmers. Interventions are therefore needed for older farmers—particularly those aged 40–50 years—to enhance these three variables. Although younger farmers show higher quality in these areas, their numbers remain limited. As such, efforts to increase young people’s interest in farming are crucial. Creating a bottom-up innovation environment may be one effective strategy for fostering such interest.

The decline in reading literacy skills among Indonesian students, seen through several international assessments such as PIRLS and PISA [1], [2], [3]], has prompted the Ministry of Education to launch various initiatives and changes in the education system to improve these skills. The results of global evaluations consistently place Indonesia among the ten countries with the lowest results in reading ability, indicating that Indonesian students face significant challenges in understanding and analyzing information from reading [4], [5], [6]. This situation indicates that serious action is needed in education reform, not only in terms of learning and curriculum but also in terms of infrastructure support. This condition requires policymakers and educators in Indonesia to design more strategic and sustainable educational interventions to address the diversity and complexity of literacy barriers students face.

The present work investigates numerous types of affine connections defined on differentiable manifolds, with a focus on Lorentzian para-Sasakian (LP-Sasakian) manifolds with non-Levi-Civita structures. We construct explicit equations for the related torsion, curvature, Ricci tensor, and scalar curvature by adding a semi-symmetric non-metric connection that is consistent with the underlying LP-Sasakian structure, illuminating their differences from traditional Riemannian equivalents. The paper rigorously describes generalized pseudo Ricci symmetric, generalized Ricci-recurrent, semi-pseudo symmetric, and semi-pseudo Ricci symmetric manifolds using covariant differential constraints on curvature and Ricci tensors. Several structural theorems are established utilizing tensorial identities, Bianchi-type relations, and contraction techniques, revealing that generalized Ricci-recurrent LP-Sasakian manifolds admitting Codazzi or cyclic type Ricci tensors necessarily reduce to Einstein manifolds under the considered connection.Furthermore, the non-existence of semi-pseudo symmetric and semi-pseudo Ricci symmetric LP-Sasakian manifolds (for dimension 𝑛>3n>3) permitting a semi-symmetric non-metric connection is convincingly shown, underlining inherent geometric impediments. The provided conclusions not only unify several ideas of curvature symmetry within a generalized affine framework, but they also provide a better understanding of how non-metricity and torsion impact the global geometric behavior of differentiable manifolds. This study adds to the larger theory of non-Riemannian geometry and establishes a foundation for future research in extended geometric structures and mathematical physics.

This study addresses a key challenge in modern networks corresponding to finding the optimal path. We first examined classical algorithms like Floyd–Warshall and Ford–Fulkerson but found them limited in scalability, flexibility, and real-time adaptability. To overcome these constraints, we developed an adaptive routing method namely ACASPO (Adaptive Cost-Aware Shortest Path Optimization) that integrates real-time updates and intelligent navigation. Simulations across diverse network structures show our approach delivers greater path efficiency with lower computational cost, particularly in large-scale or dynamic settings. Our work provides a practical, context-aware framework that connects classical theory with contemporary needs for responsive and efficient routing.

Breast cancer remains one of the leading causes of mortality among women worldwide, making early and accurate detection through mammography essential for improving patient outcomes. In recent years, Convolutional Neural Networks (CNNs) have demonstrated strong performance in automated mammogram classification tasks. However, their effectiveness is often limited to the datasets on which they are trained, as variations in imaging devices, acquisition protocols, and patient demographics introduce significant domain shift. This results in reduced performance when models are applied to unseen datasets, posing a major challenge for real-world clinical deployment. This study addresses the problem of cross-dataset generalization in CNN-based mammogram classification by incorporating domain adaptation techniques. A hybrid domain adaptation framework is proposed, integrating feature-level alignment, adversarial learning, and batch normalisation adaptation to reduce the distribution discrepancy between source and target domains. The model is trained using labelled data from a source dataset and unlabeled data from a target dataset to learn domain-invariant feature representations. Experimental evaluations conducted on multiple publicly available mammography datasets, including DDSM, MIAS, and INbreast, demonstrate that the proposed approach significantly improves classification performance across different domains. The results show notable gains in accuracy and Area Under the Curve (AUC) compared to baseline CNN models without domain adaptation. The findings emphasise the importance of domain adaptation in enhancing the robustness and generalizability of deep learning models for medical imaging, supporting the development of reliable and scalable computer-aided diagnosis systems for breast cancer detection.

The quick digitalization of the Healthcare Information System (HIS) has facilitated the efficiency of clinical activities and the accessibility of data, but has also heightened the risks to the advanced cyber-attacks like ransomware, DDoS attacks, and data breaches. The conventional intrusion detectors are not sufficient in responding to these emerging threats because they are static-rule based and centralized in design. In this paper, the researcher suggests a hybrid, safe, and interpretable intrusion detection model that combines Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), Autoencoders, and XGBoost into a weighted ensemble model. In order to save data privacy, a type of federated learning is used that allows the inclusive training of models without providing sensitive healthcare data to other participants. Moreover, explainability by SHAP is also included in order to increase transparency and trust in model decisions. CICIoT2023 and CICIDs2018 datasets were used to test the proposed model. The results of the experiments are better, as they are up to 99.99% accurate on binary classification and 99.76% accurate on multi-class IoT intrusion detection. Random Forest (99.48%), XGBoost (99.33%), and LSTM (98.51) were compared to have lower performance as baseline models. The findings also prove the claim that federated learning is as accurate as centralized training with only slight degradation. In general, the suggested framework offers a scalable, strong, privacy-sensitive solution to the security of the current healthcare infrastructures.

This paper to prevent irreversible vision impairment, it is critical to identify retinal diseases as soon as possible. At this time, most machine learning methods have been designed to identify a single retinal disease - diabetic retinopathy, glaucoma or age-related macular degeneration (AMD). Therefore, these systems are not optimal for complete disease screening of multiple retinal disorders. To address this challenge, our work proposes a supervised hybrid deep learning model that combines Convolutional Neural Networks with Vision Transformers, creating four-class fundus image classification - Normal, Diabetic Retinopathy, Glaucoma, and Age-Related Macular Degeneration. The CNN creates a localized representation of lesion characteristics such as micro aneurysms, hems, exudates, optic cup morphology, and drusen, while the VIT produces a higher-level global classification of characteristics such as disc shape, vessel patterns, and macula texture. By combining both types of characterization, our hybrid system resolves the limitations of the traditional single-disease and CNN-only networks, and will provide better performance through improved accuracy, sensitivity, and specificity as a fast, automated, and dependable method for multi-disease screening of retinal disease in clinical practice.

Human Action Recognition (HAR) is an important subfield of computer vision that has been motivated by its broad application in intelligent surveillance, human health, athletics, and human-computer interaction. Although enormous improvements have been achieved based on convolutional and recurrent neural networks, in current methods, it is frequently not practical to grasp long-range temporal structure and intricate spatio-temporal connections in video data. This paper will solve these shortcomings by making suggestions of a Multi-Attention Transformer Framework to ensure effective and powerful human action recognition. The given model combines several types of attention mechanisms, i.e., spatial attention, temporal attention, and channel attention, with the help of a transformer-based model to learn better represent features and consider global contextual dependencies. The hybrid approach to feature extraction is used, which involves the use of both convolutional layers to produce an encoding of the local spatial feature and transformer encoders to produce a long-range coding of the temporal feature. The framework is geared to efficiently deal with issues including occlusions, variations in viewpoints, and intricate movements. The performance of the proposed approach is experimentally assessed on benchmark datasets, and it is shown to be superior to the traditional CNN-based, RNN-based, and standard transformer models in terms of accuracy, precision, and computation efficiency. Multi-attention modules offer great gains to the discriminative strength of the model, without introducing new challenges to scaling the model to real-world situations. The results demonstrate the usefulness of multi-attention transformer architectures to state-of-the-art in human action recognition. The suggested framework would lead to the creation of more advanced and resilient systems of HAR, which would be a step toward further studies in multimodal and real-time action recognition.

Highly sophisticated computational modeling has become a revolutionary facilitator in the modernization of agri-tech and intelligent food processing systems, which provide intelligent, information-based solutions to global issues of food security, environmental decline, resource underuse, and losses after harvesting. With the combination of machine learning, predictive simulations, digital twins, IoT-based sensing structures, and optimization algorithms, computational modeling allows tracking the state of crops, predicting variability in yields in real-time, controlling water-energy-nutrient nexus, and identifying plant stress or disease at an early stage. The models can be used in the food processing setting to facilitate automation, quality evaluation, detection of contamination, traceability of the supply chain, as well as the development of sustainable operations with reduced energy usage. Combination of high-performance computing, sensor fusion and AI-based analytics will help develop smart, autonomous and climate-resilient agricultural ecosystems that are able to produce high productivity with minimum environmental impact. Besides, computational models are more effective in decision-making when farmers, processors, and policymakers need to simulate the results of different scenarios involving uncertainty about climate variability, resource availability, and sustainable indicators. The paper examines the purpose of state-of-the-art computational modeling as a strategic protocol of sustainable agri-tech innovations and smart food processing, its applications, advantages, implementation procedure, and the prospects in assisting the international shifts to sustainable and resilient food systems that are safe and more efficient.

The emergence of AI-driven organizational change has become a revolutionary change that allows firms to redesign their strategic models, structural designs, and decision-making system to attain sustainable competitive advantage in the ever-changing business landscape. With the development of the artificial intelligence platforms, including machine learning, predictive analytics, natural language processing and intelligent automation, companies are embracing the capabilities to develop strategic agility, provide more efficient processes, augment innovation pipelines, and remodel value creation mechanisms. The AI-enabled transformation models assist firms in mating the old top-down hierarchy models with data-based adaptive and learning ecosystems that can respond swiftly to the evolving markets. These frameworks also help to promote on-going enhancement, evidence-based decision-making, cross-functional teamwork, and allow companies to match the strategic goals with environmental, social, and governance (ESG) requirements. Through the integration of AI in the strategic planning, resource allocation, human resource development, and customer interface, organizations will be able to drive faster to be digitally mature and develop resilient systems that ensure sustained competitiveness. Nonetheless, AI-based transformation also presents threats of complexity, skills and expertise, ethical issues and aversion to change. In this paper, the strategic transformation models that are driven by AI are discussed, the contribution made to the organizational evolution is examined, the contribution made to sustainable competitive advantage is evaluated, and a conceptual framework where AI capability, strategic flexibility and sustainability-based value creation are integrated is provided.

Artificial Super Intelligence (ASI) systems need scalable and efficient decision-making models for effective functioning in complex and dynamic environments. However, existing reinforcement learning approaches have problems such as high computational complexities, low convergence rates, and poor coordination among multiple agents. Thus, this paper proposes a novel Hierarchical Multi-Agent Reinforcement Learning framework coupled with hierarchical multi-agent system architecture. The proposed framework utilizes knowledge about the environment to minimize the search space and stability in learning. Additionally, the hierarchical architecture improves coordination among agents at global, intermediate, and local levels. The system is tested using simulation techniques under artificial environment conditions. The proposed method is implemented using PYTHON software. The experimental results show that the proposed framework improves scalability by 65% to 97%, convergence rate by 30% to 98%, and computational efficiency by 60% to 92%. The experimental results validate that the proposed framework improves scalability and decision-making performance in intelligent systems. This paper emphasizes the benefits of using physics-informed learning coupled with hierarchical multi-agent reinforcement learning for future ASI applications.