Agriculture is vital to India’s economy, employing a large portion of the population, yet it faces significant challenges, such as crop diseases that cause substantial financial losses. Advanced technologies, including the Internet of Things (IoT) and cloud computing, offer effective solutions to these challenges. This study proposes a Hybrid Weighted Particle Swarm Optimization–based Faster Mask RCNN (HWPSO-FMRCNN) framework to improve the accuracy and efficiency of plant disease detection while enabling secure cloud-based analysis. Existing methods often suffer from limitations in accuracy, computational efficiency, and data security, which can hinder timely and effective plant disease management in agricultural practices. The proposed approach combines the optimization capability of weighted PSO with the feature extraction and segmentation strengths of FMRCNN for enhanced disease classification. The hybrid deep learning model analyses real-time plant images captured by IoT-enabled image acquisition devices. Cloud infrastructure provides secure storage and large-scale analysis, ensuring scalability and accessibility for farmers. By refining FMRCNN hyperparameters through HWPSO, the model achieves superior detection performance. Experimental results show that HWPSO-FMRCNN outperforms existing approaches, achieving accuracy of 98.3

In the realm of surging e-commerce data and increasing digital transactions, ensuring end-to-end security is paramount to protect sensitive information and maintain user trust. The Advanced Encryption Standard (AES), a cornerstone of network security, relies on symmetric key encryption for secure communication. However, rapid advancements in cryptanalysis and computational power pose new threats to AES, particularly exposing vulnerabilities related to its static S-Box design. This article is motivated by the urgent need to enhance cryptographic resilience in dynamic threat environments, especially for applications involving online financial systems, secure cloud storage, and real-time communication. To address these concerns, we propose an involution-based concurrent error detection (CED) mechanism for dynamic twin-key generation, significantly improving resistance against differential and linear cryptanalysis. By exploring the evolution of involution structures in stream cipher construction, we introduce a low-latency CED framework tailored for symmetric block ciphers. The proposed method outperforms traditional AES-128 by leveraging a dual-key encryption approach. Additionally, we design a programmable, key-dependent S-Box architecture that enhances unpredictability and fortifies the encryption process, particularly in hardware-constrained environments such as IoT devices and embedded systems. Mathematical analysis substantiates the robustness and practicality of the proposed enhancements, offering a future-ready solution to secure modern data systems against emerging decryption techniques.

This study introduces a novel approach to sustainable fuel production through the integrated application of plasma gasification, Fischer-Tropsch (FT) synthesis, and solar-powered electrolysis, focusing on the covalorization of medical waste (MW) and biomass waste (BMW). Addressing critical challenges related to waste accumulation and energy security, the system transforms complex waste streams into synthetic e-fuels by optimizing syngas composition and hydrogen integration. In the case study of Iraq where over 4.12 million tonnes of solid waste were landfilled in 2023, including substantial medical and municipal waste this approach offers a viable solution for resource recovery. Plasma gasification converts MW and BMW into syngas, which is refined and processed for CO2 capture using an MDEA-based absorption system. Solar-derived hydrogen, produced through a PEM electrolyzer, is integrated to achieve the optimal H2/CO ratio for FT synthesis. System modeling was conducted in Aspen Plus and MATLAB, with a Genetic Algorithm (GA) employed to optimize parameters for hydrogen yield enhancement. Four biomass-to-medical waste blending ratios (0.2 to 0.8) were tested. The highest-performing scenario (0.8 ratio) achieved a hydrogen mole fraction of 45.78 %, a syngas flowrate of 8,670 Nm3/h, hydrogen production of 1,300 kg/h, and a peak FT conversion efficiency of 55.8 %. Liquid fuel yield reached 1,360 kg/h, with diesel comprising the dominant product at 519 kg/h and 22,317 MJ/h of energy output. Hydrogen utilization efficiency increased to 87.5 %, and energy cost for hydrogen electrolysis decreased to 54.2 MJ/kg. Economic evaluation revealed strong financial viability at scale, with the highest scenario yielding a net present value (NPV) of $43.71 million, return on investment (ROI) of 16.48 %, and a payback period reduced to 14.93 years. Environmental analysis showed significant reductions in CO2 emissions (down to 19.3 %) and improvements in carbon-to-fuel efficiency (up to 71.2 %), with stable solid residue losses.

Assessing sleep quality is essential to preserving optimum health and well-being, with consequences ranging from preventing chronic diseases to improving cognitive function. This paper introduces a sophisticated hybrid deep learning architecture that far outperforms current techniques for actigraphy data-based sleep quality prediction. Our method uses two metaheuristic optimization approaches (genetic algorithms and particle swarm optimization (PSO)) for feature selection and combines statistical characteristics with complex features retrieved using long short-term memory (LSTM) networks. support vector machines (SVMs) are then used to classify the optimized feature set. Our model outperforms baseline LSTM and other cutting-edge methods when tested on the benchmark MESA Actigraphy dataset. It achieves remarkable accuracy (84.64% for weekly sleep quality and 68.99% for sleep consistency), F1-scores (0.847 and 0.69, respectively), and AUC values (0.909 and 0.839, respectively). Furthermore, we close a significant gap in black-box deep learning techniques by introducing a unique feature significance analysis that gives the model's predictions interpretability. Our results emphasize the potential of hybrid deep learning frameworks for individualized sleep health management and early diagnosis of sleep disorders by demonstrating the efficacy of integrating metaheuristic optimization with multimodal data in sleep quality prediction.

The integration of technology into educational institutions has led to the generation of vast data, creating opportunities for Educational Data Mining (EDM) to improve learning outcomes. This study introduces a novel feature selection model, “Dynamic Feature Ensemble Evolution for Enhanced Feature Selection” (DE-FS), which combines traditional methods such as correlation matrix analysis, information gain, and Chi-square with heat maps to select the most relevant features for predicting student performance. The core innovation of DE-FS lies in its dynamic and adaptive thresholding mechanism, which adjusts thresholds based on evolving data patterns, addressing the limitations of static methods and mitigating issues like overfitting and underfitting. This research makes three key contributions: it introduces an advanced ensemble-based feature selection methodology, incorporates dynamic and adaptive thresholding to improve accuracy and flexibility, and demonstrates DE-FS’s superior predictive performance across diverse educational datasets. The results highlight DE-FS’s ability to adapt to fluctuating data patterns, enabling precise and reliable student performance predictions, supporting targeted interventions, and improving resource allocation to enhance personalized learning experiences.

PURPOSE:This study presents NeuroDL, a novel deep learning-based diagnostic framework designed for the automated detection of brain tumors and Alzheimer's disease (AD) using magnetic resonance imaging (MRI). The objective is to enhance diagnostic precision and efficiency in neurology through advanced computer-aided decision support. METHODS:NeuroDL utilizes convolutional neural networks (CNNs) trained on two publicly available, annotated MRI datasets. The proposed pipeline integrates optimized preprocessing, including normalization, skull stripping, and data augmentation, followed by CNN-based feature extraction and classification. Transfer learning and fine-tuning were employed to improve generalization on limited medical data. RESULTS:Experimental evaluations show that NeuroDL achieves 96.8% classification accuracy for brain tumor detection and 92.4% accuracy for Alzheimer's disease diagnosis. The method also achieves an average F1-score of 0.965, precision of 0.969, and recall of 0.962 for brain tumors, and an F1-score of 0.918, precision of 0.921, and recall of 0.916 for AD. These results outperform state-of-the-art benchmarks on the same datasets. CONCLUSION:Potential for real-time clinical deployment. It addresses key limitations of existing CAD systems by providing a unified, dual-disease diagnosis model with statistically validated performance gains. NeuroDL paves the way for reliable, scalable, and automated neurological disease diagnosis using deep learning.

Data clustering is a fundamental technique in data mining, pivotal for various applications such as statistical analysis and data compression. Traditional clustering algorithms often struggle with noisy or high-dimensional datasets, hindering their efficacy in addressing real-world challenges. In response, this research introduces MutaSwarmClus, a novel hybrid metaheuristic algorithm that combines Mouth Brooding Fish (MBF), Ant Colony Optimization (ACO), and mutation operators to enhance clustering quality. MutaSwarmClus intends to adaptively control the exploration and exploitation phases of the solution space, solve issues with local optima and changes in the distribution of available data. Moreover, it incorporates an Iterated Local Search (ILS) to refine solutions and avoiding getting stuck in local optima. MutaSwarmClus therefore increases the robustness of the clustering process by incorporating controlled randomness through mutation operators in order to handle noisy and outlier data points well. According to the contributions analysis, the proposed algorithm improves the clustering solution with the combined system of MBF, ACO, and mutation operators, which enables the mechanism of exploration and exploitation in the process of information search. As shown through the results of experimental studies, MutaSwarmClus has high performance when used with various benchmarks, and outperforms or performs as well as or better than compared to other clustering algorithms such as K-means, ALO, Hybrid ALO, and MBF. It achieves an average error rate of only 10%, underscoring its accuracy in clustering tasks. The utilization of MutaSwarmClus offers a solution to the existing problems in clustering large datasets in terms of scalability, efficiency and accuracy. Possible directions for future work can continue to optimize the model parameters of the algorithm and study its adaptability in dynamic conditions and with large amounts of data.

Higher education is essential because it exposes students to a variety of areas. The academic performance of IT students is crucial and might fail if it isn’t documented to identify the features influencing them, as well as their strengths and shortcomings. The student academic prediction system needs to be enhanced so that teachers can forecast their students’ performance. Numerous studies have been conducted to increase the prediction accuracy of IT students, but they encountered difficulties with unbalanced data and algorithm tuning. To address these issues, the study proposed different machine learning (ML) algorithms that handled imbalanced data by applying the synthetic minority oversampling technique (SMOTE) and employing hyperparameter tuning algorithms to enhance prediction during the training process. The ML models we used were decision tree (DT), k-nearest neighbor, and XGBoost. The models were fine-tuned by applying Ant colony optimization (ACO) and artificial bee colony optimization techniques. Subsequently, these optimization techniques further enhanced the performance of the models. After comparing them, the results showed that SMOTE and ACO combined with the DT model outperformed other models for academic prediction. Additionally, the study utilized the Kendall Tau correlation coefficient technique to analyze the correlation between features and identify factors that positively or negatively impact student success.

This study investigates the effectiveness of Abrasive Water Suspension Jet (AWSJ) Machining, a non-conventional erosion-based method, for machining carbon fiber-reinforced polymer (CFRP) composites. The focus was on analyzing key process parameters—abrasive size, feed rate, and standoff distance (SOD)—under submerged cutting conditions and their impact on material removal rate (MRR), kerf width, and surface roughness. Experimental trials were conducted, and advanced computational techniques, including Response Surface Methodology (RSM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Artificial Neural Networks (ANN), were used for parameter optimization and predictive analysis. The results showed that submerged cutting significantly improved machining quality by reducing surface roughness and ensuring uniform kerf widths. Increasing the jet diameter in underwater conditions stabilized the nozzle, leading to smoother and more precise cuts. Among the predictive models, XGBoost demonstrated the highest accuracy and efficiency in forecasting MRR, while Random Forest and ANN provided competitive performance. The integration of RSM and machine learning (ML) techniques enabled effective optimization of machining parameters, showcasing the potential for cost-effective and high-precision CFRP machining. These findings are particularly relevant for industries like aerospace and automotive, where machining efficiency and precision are crucial.

Thyroid cancer recurrence poses a significant challenge in clinical practice, often complicating treatment outcomes and long-term patient management. This study investigates the application of multiple machine learning models—Logistic Regression, Decision Trees, Random Forest, Gradient Boosting, Support Vector Machines (SVM), and K-Nearest Neighbors (KNN)—to predict recurrence using a real-world dataset of 383 patients. Among the evaluated models, the Random Forest classifier achieved the highest accuracy of 98.26%, demonstrating a strong balance between sensitivity and specificity. To mitigate class imbalance, class weights were incorporated into the model training. Model performance was assessed through Stratified 5-Fold Cross-Validation. For robust hyperparameter tuning and a reliable performance estimate, we implemented nested cross-validation with GridSearchCV. Analysis of learning curves provided insights into the model’s behavior with varying training data sizes. Furthermore, we assessed and improved the reliability of the model’s predicted probabilities through model calibration using reliability curves and Isotonic Regression, which is crucial for clinical decision-making. The model demonstrated promising performance (mean nested CV accuracy: ~0.964), and the calibration significantly improved the trustworthiness of probability estimates. Comprehensive preprocessing and feature selection methods, including Chi-square and Random Forest Gini Importance, were applied to enhance model performance. Further explainability was added using SHAP analysis to understand key feature contributions. The findings underscore the potential of machine learning-based frameworks to support early identification of high-risk recurrence cases, thereby assisting clinicians in tailoring follow-up and treatment strategies more effectively.

Population-based metaheuristic optimization algorithms have gained prominence for tackling complex optimization problems. They balance exploration and exploitation, essential for finding optimal solutions. While algorithms like Genetic Algorithms, Particle Swarm Optimization, and Gravitational Search Algorithm have shown success, they have limitations, such as premature convergence and sensitivity to parameters. To address these issues, we have introduced Quantum-Inspired Gravitationally Guided Particle Swarm Optimization (QIGPSO) for addressing complex optimization challenges, particularly in the context of medical data analysis for diagnosing Non-Communicable Diseases (NCDs). The Quantum Particle Swarm Optimization (QPSO) and Gravitational Search Algorithm (GSA) are both used in QIGPSO. It takes advantage of each algorithm’s strengths in both global and local search processes. We used an absolute Gaussian random variable to improve the search, changed the position update equations and used a wrapper-based method with Support Vector Machine (SVM) for feature selection and classification. The findings suggest that QIGPSO is effective at identifying key features, achieving high accuracy rates, and lowering the number of incorrect classifications across several NCD datasets. Doctors can use QIGPSO data to make better treatment decisions for their patients. QIGPSO overcomes the limitations of conventional optimization methods by faster convergence while improving exploitation balance.

In our data-driven world, the healthcare sector faces significant challenges in the early detection and management of Non-Communicable Diseases (NCDs). The COVID-19 pandemic has further emphasized the need for effective tools to predict and treat NCDs, especially in individuals at risk. This research addresses these pressing concerns by proposing a comprehensive framework that combines advanced data mining techniques, feature selection, and meta-heuristic optimization. The proposed framework introduces novel hybrid algorithms, including the Hierarchical Genetic Multiple Reduct Selection Algorithm (H-GMRA) and the Customized Function-based Particle Swarm Optimization with Rough Set Theory for NCD Feature Selection (CPSO-RST-NFS). These algorithms aim to address the challenges of feature selection, computational complexity, and disease classification accuracy. H-GMRA outperforms traditional methods by identifying minimal feature sets with high dependency ratios. CPSO-RST-NFS combines meta-heuristic optimization with feature selection, resulting in improved efficiency and accuracy. Through extensive experimentation on diverse NCD datasets, this research demonstrates the framework’s ability to select informative features, improve classification accuracy, and contribute to better patient outcomes. By bridging the gap between computational efficiency and disease classification accuracy, this work offers valuable insights for healthcare practitioners and data analysts, ultimately advancing the field of NCD research. The proposed framework presents a significant step towards enhancing the early detection and management of NCDs, offering hope for more precise clinical predictions and improved patient care.

Driver fatigue is one of the most common causes of road accidents, which means that there is a great need for robust and adaptive monitoring systems. Current models of fatigue detection suffer from domain-specific limitations in generalizing across diverse environments, sensor variability, and individual differences. Moreover, they are not resilient to real-time sensor quality issues or missing data, which limits their practical applicability. To overcome the aforementioned challenges, we propose a holistic Dynamic Cross-Domain Transfer Learning framework for fatigue monitoring application using multi-modal sensor data fusion. There are four innovations involved with this framework. Firstly, the domain adversarial neural network in EEG, ECG, and video inputs ensures cross-domain invariance of features. The gap of adaptation at the domain goes below 5%, while there is an improvement of the cross-domain accuracy to as high as 15% from 10%. The ASF-Transformer uses adaptive cross-modal attention for fusing heterogeneous sensor data effectively. Accuracy improves by 5-8% and remains robust under modality dropout conditions. Third, the GMSN dynamically evaluates sensor quality and selectively enables modalities to mitigate performance drops to < 5% even with noisy or missing inputs in process. Fourth, Online Personalized Fine-Tuning (OPFT) allows for real-time adaptation of the model to individual drivers, achieving an improvement in accuracy by 5-7% within 2 h with a latency of < 50ms. Thorough evaluations show that the framework can achieve 85-90% accuracy on target domains while maintaining robustness under 20% sensor dropout. Addressing the issue of domain variability, sensor quality, and personalization, this work has improved the reliability, adaptability, and real-time feasibility of fatigue monitoring systems to provide significant advancements for driver safety in dynamic real-world environments.

Urban grids face the challenge of expanding renewable deployment while curbing emissions and minimizing the capital burden of network reinforcements, all of which depend on effective flexibility integration. A hybrid optimization framework is introduced, combining Mixed-Integer Nonlinear Programming with a reformulated MILP structure to jointly size photovoltaic systems, battery storage, staged network upgrades, and the dynamic participation of electric vehicles as both load and distributed storage. Thousands of EV constraints are consolidated through a polytope-based approach, and reinforcement costs are captured using a piece-wise linear model tailored to feeder capacity increments. Application to the Tuwaiq Smart City network, covering 3780 households, one-minute resolution data for 2024 to benchmark five schemes: No Demand Response, Smart Charging, Vehicle-to-Grid, and Integrated Decentralised Energy Management (IDEM). Compared with the baseline, IDEM achieves a 43.8 % reduction in annualised system cost, 46 % decrease in peak imports, and capacity cuts of 75 % and 82 % for PV and storage respectively, alongside a 65 % drop in grid integration expenses. A Monte Carlo test of 150 runs confirms cost stability within +/- 6 %, validating the robustness of layered flexibility under stochastic solar and mobility profiles. Solving across a full-year span is achieved within minutes on standard hardware, confirming the framework's practical value for strategic energy planning

This paper explores the use of industrial sludge as a renewable energy source, addressing challenges in sustainable energy production and waste management. It investigates the varied composition and properties of sludge from sectors like wastewater treatment, power generation, and chemical production, highlighting its potential for energy recovery. The paper provides an elemental analysis to underscore the energy content of different sludge types and examines methods for converting sludge into viable energy resources. Techniques such as anaerobic digestion, pyrolysis, gasification, and advanced thermal and electrochemical conversions are evaluated for their feasibility, environmental impact, and economic implications. The research emphasizes optimizing these methods for integration into existing energy systems and the importance of comprehensive cost and environmental impact assessments. It identifies main challenges in sludge conversion, including high moisture content and the presence of contaminants, underlining the need for effective management strategies. Future research directions proposed include diversifying across industries, refining energy extraction methods, conducting thorough environmental assessments, and developing supportive policies for this innovative approach. The paper concludes with a call for collaborative stakeholder efforts to advance this sustainable energy paradigm, highlighting the significant contribution of harnessing industrial sludge's energy potential for a more sustainable and environmentally friendly future.

Cognitive Radio Ad-hoc Networks (CRAHNs) are under constant attacks from compromised primary & secondary nodes. These attacks focus on bandwidth manipulation, internal configuration manipulation, and selective spoofing, which can disturb the normal working of the CRAHNs. Researchers propose various security models to mitigate these attacks, each with limitations. Most of these models have higher complexity, while others cannot be used to mitigate multiple attack types. To overcome these issues while maintaining higher security and Quality of Service (QoS) under attacks, this text proposes a design of a novel blockchain-based security model for improving attack resilience in CRAHNs. The model initially collects multiple information sets from different cognitive radio controllers and creates active & redundant miners for the storage of these sets. The number of active & redundant miners is decided via a Mayfly Optimizer (MO) Model, which assists in improving resource utilization while reducing deployment costs. Cognitive rules and configurations are stored on these nodes and updated via a secure blockchain verification. Due to this, the proposed model demonstrated significant improvements in cognitive radio communications across various metrics, even under different attack scenarios. It reduced communication delay by up to 18.5%, increased communication throughput by up to 19.5%, and improved the Packet Delivery Ratio (PDR) by up to 19.4% when compared with existing models such as SRC, Prob Less, and DDQL. Additionally, the model achieved energy savings of up to 12.5%. These enhancements were made possible by the optimized selection of miner nodes, enabling quicker mining for high-speed communication, low-energy mining tasks for prolonged use, and high-performance mining for consistency. The results affirm the model’s suitability for various real-time cognitive radio scenarios. Due to the integration of the MO Model, the CRAHN showcases better communication speed, lower energy consumption, higher throughput, and higher packet delivery performance when compared with existing methods under real-time scenarios.

The rapid increase in the concentration of carbon dioxide (CO2) polluting the atmosphere induces global warming and climate change. This is detrimental to human health and their natural habitat. Thus, it is imperative to proffer measures in analyzing and predicting the emissions of CO2. This research suggests using an ensemble approach with fuzzy nearest neighbor, sequential minimal optimization, and logistic regression to predict global CO2 emissions. The K-means algorithm divides data into groups of similar and relevant patterns. Simulation findings show that the proposed model outperforms techniques such as multi-layer perceptron, fuzzy ownership nearest neighbor, and random forest. It also improves CO2 forecast accuracy.

Using Density functional theory (DFT), we investigated the charge storage capacity, quantum capacitance (C-Q), geometry and electronic structures of BNyne/graphene heterostructures (BNyneGHs), as well as the impact of transition-metal dopants on their C-Q. Our results showed that doping was more effective than vacancy defects in improving the C-Q of BNyneGHs. Ti-doped BNyneGHs exhibited the highest C-Q value of 360.08 mu F/cm(2), making them ideal positive electrode materials for supercapacitors (SCs). The presence of doping agents was found to enhance the density of states (DOS) around the Fermi level, resulting in improved C-Q. Our calculations identified potential cathode or anode materials for high-energy-density SCs, providing theoretical support for the design of high-capacitance SCs.

Today, designing a green hydrogen production process under an optimal, efficient, and economical configuration is one of the priorities of energy systems engineers. This research aims to explore the application of electricity produced by the steam Rankine cycle (S/RC) and Organic Rankine cycle (ORC) through biomass-based waste heat utilization in an alkaline electrolyzer (AE) system for the production of green hydrogen. The S/RC unit utilizes high-temperature waste heat, while the ORC system utilizes medium to low-temperature ones from the S/RC to improve overall system efficiency and hydrogen output. The study developed eight distinct configurations of the combined system, employing two organic fluids across four ORC designs. It aimed to assess the influence of integrating a recuperator or/and an OFOH (open feed organic heater), as well as varying the organic fluids, on the ORC unit's hydrogen production capability. A detailed thermodynamic, thermo-economic, and eco-economic analysis was conducted to assess key metrics. The environmental analysis quantified the carbon dioxide (CO2) emission reductions achieved through hydrogen production, using the emission rate from the steam methane reforming method as a benchmark. This approach underscored the environmental benefits of the AE system for hydrogen production. Further, the study quantified the financial gains from CO2 reduction, referred to as carbon credit gain (CCG), through an eco-economic analysis. From the outcomes, the highest hydrogen yield and the lowest LCOH (levelized cost of hydrogen) value were 94.65 tons/year and 1.724 US$/kg, respectively, related to Case (IV)-a (biomass waste heat-based S/RC- Regenerative& recuperator ORC system-AE system under R245fa). Lastly, optimization process for maximizing hydrogen production via the optimization methodology was established.