Efficient alignment between power demand and supply is essential for sustaining a dependable and sustainable energy system. Maintaining the equilibrium is essential for creating resilient energy management systems that can satisfy the increasing needs of contemporary civilization. This current study proposes an innovative predictive model that utilizes Explainable Artificial Intelligence (XAI) to improve energy forecasting. The model, based on the eXtreme Gradient Boosting (XGBoost) method, incorporates Shapley Additive Explanations (SHAP) to enhance interpretability by emphasizing the impact of critical parameters on energy consumption forecasts. Historical electricity consumption data is processed using feature engineering to create temporal and weather-related features, and sliding-window approach has been adopted to capture temporal dependencies. The proposed model is assessed using a real-time home electricity dataset, attained mean absolute error (MAE) of 0.119 kWh and mean squared error (MSE) of 0.0321 kWh while offering explainable insights about produced electricity predictions. The explainable model offers understandable electricity demand estimates, hence aiding in the development of energy management systems that are efficient and robust to rising demand.
E-commerce platforms generate large volumes of user feedback; however, extracting meaningful sentiment information for accurate and personalized recommendations remains challenging due to noisy textual data and limitations in existing models. To address this issue, this study proposes a novel Multiagent AI-based Recommender System integrating advanced feature extraction, optimized sentiment classification, and fuzzy multiagent reasoning. The framework employs the aggregated backbone network (ABNet) for semantic feature representation and an optimized cascaded attention transformer network (CATNet) for context-aware sentiment classification, with hyperparameters tuned using the enhanced wombat optimization algorithm (EWOA). Final recommendations are generated through a fuzzy belief multiagent AI recommendation (FBMAR) mechanism. Experimental evaluation on benchmark datasets demonstrates that the proposed CATNet achieves a sentiment classification accuracy of 98.25
In a sustainable hybrid energy system, this research suggests optimizing hybrid energy systems using renewable energy sources (RES). The behaviour of renewable energy is ambiguous, making static optimization challenging ways to make the hybrid system’s uncertain non-stationary distributed energy resources as efficient as possible. For the system of hybrid energy, a stochastic-based multi-objective strategy is developed to optimize overall system losses and operational costs. The suggested objective function is to reduce system losses and the overall operating expense of RES (solar and wind) across the power station. Here, a new model of load demand and network associated RES is proposed while taking uncertainty and variability into account. By employing the distribution functions of various probabilities that are used to describe the RES statistics, a robust stochastic technique is here given. The system operations under uncertainty are handled by the simulation results in this research. On the distribution system of the IEEE 37 node, the proposed approach has been tested. The simulation results demonstrate the hybrid energy system’s effectiveness in the suggested optimization strategy as compared to the IEEE 33 node distribution system.
LNG, with its low emissions and high calorific content, has become a vital energy source recently. Cryogenic characteristics and fire and explosion dangers require stringent safety examination, especially in unloading terminals with significant inventory and dynamic operating conditions. Classic criticality and risk assessment tools like Failure Mode Effect and Critical Analysis (FMECA) are widely used, but FMECA requires precise failure data and cannot distinguish between failure modes with the same Risk Priority Numbers (RPN). This work proposes a hybrid methodology for the LNG storage facility, integrating adaptive neuro-fuzzy inference system (ANFIS), fuzzy logic, and expert elicitation to address these constraints. Expert opinions from design, operation, and maintenance professionals were quantified using Triangular Fuzzy Number (TFN)–based fuzzy membership functions to translate linguistic variables into numbers. Fuzzy inference was used to estimate uncertainty and improve prioritizing using fuzzified severity (S), occurrence (O), and non-detection (D) inputs. In addition, ANFIS optimized Gaussian membership functions and inference procedures to create discriminative and continuous RPN scores. The fuzzy–ANFIS FMECA model outperforms classic FMECA in risk ranking precision, stability, and resolution, making it a useful LNG storage facility design decision-support tool.
The increasing frequency of natural disasters, geopolitical conflicts, pandemics, and supply chain disruptions has exposed the vulnerability of logistics and transportation networks, making resilience a critical priority for organizations and policymakers. Despite the growing body of research on logistics resilience, limited attention has been given to understanding how resilience-enhancing strategies can be prioritized and implemented under resource constraints while accounting for the trade-offs among competing objectives. This study aims to address this gap by investigating whether strengthening logistics resilience is fundamentally a matter of strategic trade-offs and by proposing an integrated decision-support framework for resilience planning. The study introduces a novel hybrid approach that combines the Best–Worst Method (BWM) and Goal Programming (GP) to evaluate, prioritize, and optimize resilience strategies. Twenty resilience strategies are assessed across four key dimensions infrastructure, operational, digital, and policy using expert judgments to determine their relative importance, followed by GP-based optimization to allocate limited resources efficiently under budgetary constraints. The results reveal that strategies such as dynamic routing, real-time tracking, and investments in digital infrastructure exert the greatest influence on enhancing logistics resilience. However, their implementation requires balancing trade-offs related to cost, feasibility, technological compatibility, and organizational priorities. The findings demonstrate that effective resilience planning extends beyond identifying high-impact interventions and requires systematic decision-making to reconcile competing objectives across multiple domains. By framing logistics resilience as a constrained optimization problem, this study contributes to both theory and practice by offering a comprehensive and actionable framework for resilience assessment and resource allocation. The proposed approach provides valuable insights for transport planners, logistics managers, and policymakers seeking to improve the robustness, adaptability, and sustainability of logistics systems in an increasingly uncertain environment.
As the fifth-generation (5G) wireless communication network is introduced worldwide, the expansion of Internet of Things (IoT) services continues, covering a wide area of applications, including enhanced mobile broadband, improved low-latency communication services, drone-based systems, reliable Internet connectivity with high-bandwidth capacity, and non-terrestrial networks. However, these applications may not work perfectly with 5G because 5G is mainly designed to handle small data packets and very fast, low-delay communication. To satisfy the wide-ranging requirements of IoT, researchers have proposed next-generation networks (NGNs). NGNs are not only limited to terrestrial networks and incorporate non-terrestrial components, ensuring high-reliability and high-capacity connectivity through drones, high-altitude aeronautical platforms (HAAPs), and satellites. NGNs are designed to provide seamless connectivity across both urban and remote regions by integrating diverse communication infrastructures. They support real-world applications such as intelligent transportation, smart healthcare, industrial automation, and emergency response, where reliable communication is essential. Therefore, the reliability assessment of NGNs is critical to ensuring robust, continuous, and dependable communication. Assessing the reliability of this architecture is critical to ensuring optimal network performance. In this paper, we propose a novel NGN architecture that integrates drones and HAAPs and performs a comprehensive reliability analysis. A five-level layered architecture is designed, and reliability is derived using analytical methods. Stochastic modeling approaches, such as reliability block diagrams, are used to construct the models. The graphical results show that the satellite and cloud layers are more reliable than drones and HAAPs. Also, increasing the number of drones raises the risk of failure, whereas increasing additional HAAPs improves coverage. A comparative reliability degradation of NGN network components is also performed. The analytical results are further validated using a MATLAB-based discrete-event simulation, with the simulation outcomes showing close agreement with the analytical results.
This study examines how Architecture, Engineering, and Construction (AEC) enterprises in Indonesia progress toward Level 3 Building Information Modeling, also referred to as integrated BIM (i-BIM) maturity. Using a cyclical action research design, the study engaged five enterprises across fifteen building and infrastructure projects, combining case study evidence with focus group validation. The research addressed three objectives: (1) to examine BIM adoption processes by identifying key drivers and barriers influencing organizational maturity; (2) to identify and prioritize success factors enabling progression across BIM maturity levels; and (3) to develop a practical framework integrating BIM maturity models with organizational change perspectives, tailored to developing economy contexts. The findings indicate that although technological capabilities—such as 3D–7D modeling, 4D sequencing, and Common Data Environments—were implemented, progression toward Level 3 BIM was constrained by fragmented processes, skill gaps, and cultural resistance. Nine critical success factors were validated, with process performers, process design, enterprise culture, and enterprise leadership emerging as the most influential enablers. Strengthening organizational and process capabilities was shown to be a prerequisite for realizing the benefits of advanced BIM technologies and policy instruments. The proposed framework integrates Hammer’s process and enterprise maturity areas with the BIM capability dimensions proposed by Change Agents AEC, offering structured guidance for enterprises seeking to progress from partial collaboration toward fully integrated (Level 3 BIM) practices. While grounded in the Indonesian context and a single action research cycle, the framework is adaptable to other developing economy settings.
Smart manufacturing needs energy-efficient and sustainable motion control systems to achieve better productivity results. Existing controllers which include PID controllers exhibit tracking errors and delays while consuming more energy during complex trajectory tracking and different operational conditions. The research presents a novel Hybrid Fuzzy-Kinematic Motion Optimization Controller (HFKMOC) which establishes a connection between fuzzy logic adaptive PID control and kinematic motion optimization. The novel algorithm establishes energy-efficient motion planning on the basis of real-time implementation of the trajectories and the adaptive control of the gains to decrease the position error and to make the motion smoother. The new controller tested benchmark trajectory data sets which included sinusoidal, step and non-linear trajectories to evaluate its tracking accuracy and mean position error and energy efficiency and motion smoothness and jerk reduction capabilities. The HFKMOC system outperformed all tested controllers which included conventional PID and Fuzzy PID and ANFIS and MPC and GWO-PID by achieving 98.3
Recently, one of the most powerful areas in computer science is Artificial Intelligence (AI). AI can transform the practice of medicine and the delivery of Healthcare (HC) organizations. Machine Learning (ML) and Deep Learning (DL) are subsets of AI that employ approaches for learning patterns from data. These ML and DL algorithms focus on analyzing and interpreting patterns to solve numerous challenges, like disease predictions, drug discovery, and so on, in intelligent HC applications. In intelligent HC, ML and DL algorithms mainly predict outcomes more accurately and help to solve efficient and robust diagnostics challenges. Here, Predictive Analytics (PA) is explored to overcome the challenges in HC, showing the importance of ML and DL. PA risk assessment plays a crucial role in appraising an individual’s risk of developing certain diseases. The risk assessment makes significant contributions to ML and PA for enhancing the ability to enable early detection and diagnosis of diseases. Therefore, ML, DL, and PA can reveal complex patterns and identify intelligent indicators that are complex for human clinicians. Thus, this review explains the Healthcare Services (HCS), the role of ML and DL in improving HCS, the role of PA, and HCSs’ performance with ML and DL in optimizing HCSs.
In the age of fast technological growth, maintaining a balance between privacy rights and technological innovation presents global regulatory issues. Traditional forensic techniques are unable to handle the volume and complexity of digital evidence as cybercrime gets increasingly complicated. Artificial intelligence can be quite helpful in this situation by automating labour-intensive operations, improving data processing, and producing more accurate outcomes. India’s revolutionary attitude with the Personal Data Protection Act 2023 is the main subject of this paper, which examines legislative measures across jurisdictions. The paper also examines how these frameworks strike a compromise between privacy and innovation by comparing their conformance to the California Consumer Privacy Act (CCPA) in the US and the General Data Protection Regulation (GDPR) that is widely used in the EU. The study examines the sociopolitical backdrop of India, looking at how it affects digital freedoms and democratic government. This analysis helps policymakers, legal professionals, and researchers to navigate privacy rights in the face of technological advancement by providing insights into India’s regulatory strategy and international standards. This is essential for bridging the gap between privacy laws and technological innovations and helping to shape an equitable framework in India.
Students' lack of motivation and low interest in studying is a key issue that educational institutions and educators frequently address. Advancements in technology, the Internet, and ICT tools have significantly changed the educational landscape and learning environment over time. This study aims to identify and prioritize the gamification features applicable/used in higher education to improve students' engagement with the learning process with the help of a multi-criteria decision modelling approach—the analytical hierarchy approach (AHP). This study prioritizes the five gamification features (reward, feedback, challenges, progress, and points) used by universities that provide higher education to enhance students' engagement with the learning process. The study found that higher education's most effective gamification attribute is rewards provided to students based on their performance and achievements to increase their engagement. Points follow this in the learning process, as the students are highly interested in collecting and accumulating points in the academic journey to use them for other rewards. Other influential gamification features in the educational environment include progress communication, feedback, and challenges. This study provides strategic insights for universities providing higher education courses to use the different gamification features in the learning system to increase the system's effectiveness by enhancing students' engagement with academic journeys. However, the prioritization is grounded in expert academic judgment and is subject to empirical validation through direct student data in future research.
Intrusion detection in Wireless Sensor Networks (WSNs) is crucial for ensuring network security and data integrity. This study proposes an edge-computing-based intrusion detection model using machine learning techniques applied to the WSN-DS dataset. The primary objective is to develop an effective model capable of accurately distinguishing between normal and malicious activities, particularly Denial of Service (DoS) attacks. Several machine learning algorithms were implemented and evaluated, including Random Forest, Gradient Boosting, Logistic Regression, K-Nearest Neighbors, Decision Tree, Gaussian Naive Bayes, and Multilayer Perceptron (MLP) classifiers. The proposed methodology incorporates extensive data preprocessing steps such as handling missing values, one-hot encoding for categorical features, feature standardization, and Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance. Model performance was assessed using accuracy, precision, recall, F1-score, and Area Under the Curve (AUC). Random Forest, Gradient Boosting, and Decision Tree models demonstrated excellent performance, achieving an AUC of 1.0, outperforming Logistic Regression. Gaussian Naive Bayes showed comparatively lower performance, with precision reaching approximately 93
Industry 4.0 smart manufacturing has created a need for intelligent robotic control systems to be able to successfully track the trajectory and maintain stable motion in dynamic industrial environments. However, existing PID, fuzzy PID and intelligent control methods tend to lack accuracy in tracking, have poor convergence performance and cause overshoot and are not adaptable to nonlinear disturbances, multi-axis coupling and variable operating conditions. To deal with these problems, this research presents an Adaptive Fuzzy Kinematics-Driven PID Control (AFKP-Control), which is able to enhance the accuracy of robotic motion and operation robustness. The designed control method is adaptive PID control, fuzzy logic inference, real-time kinematic modeling, IoT based sensor feedback, triangular membership functions, Mamdani fuzzy reasoning and self-tuning gain adaptation, which is able to adapt the PID control parameters in real-time according to the system error and the change of system error. The framework constantly tracks the discrepancies in the robot’s trajectory, constantly optimizes the control gains in real time, considers the velocity and acceleration of the robot and most importantly calculates optimized control actions for robots to move with accuracy. Experimental tests on a 3-DOF robotic manipulator show that the proposed scheme can achieve better trajectory tracking, disturbance rejection, control stability and energy efficiency than other intelligent control schemes. The proposed AFKP-Control framework not only provided for low complexity of implementation and fast convergence, but also demonstrated the accuracy tracking of 97.4
Student lifestyle factors play an important role in shaping academic performance, yet their relative influence remains insufficiently quantified using data-driven approaches. This study proposes a lifestyle-based predictive framework for educational outcomes using machine learning regression models integrated with sensitivity analysis techniques. A dataset comprising key lifestyle variables study hours, extracurricular activities, sleep duration, social engagement, physical activity, gender, and stress level was analyzed using Random Forest Regression (RFR) and eXtreme Gradient Boosting Regression (XGBR). Model performance was enhanced through hyperparameter optimization using the Pelican Optimizer (PO) and evaluated via 5-fold cross-validation employing multiple statistical metrics. To improve interpretability, FAST sensitivity analysis was applied to diagnose and rank the impact of input variables on model outputs. Results demonstrate that the optimized XGBR-based model achieved superior predictive accuracy and stability. Sensitivity analysis identified stress level as the most influential factor (ST = 0.910), followed by extracurricular activity hours (0.894) and gender (0.885). Negative sensitivity values indicate inverse relationships, where increases in certain inputs correspond to decreases in predicted academic outcomes, highlighting the directional nature of lifestyle effects. These findings emphasize the importance of combining robust machine learning models with sensitivity diagnostics to support interpretable and actionable educational outcome prediction.
This vmulti-case study research fills gaps in sustainable fashion practices and improves sustainable retailing understanding. Booms Bitner’s 7Ps retail mix paradigm, previously criticized for overlooking environmental factors, is reinterpreted by embracing circular economy ideas to provide a more sustainable strategic model for decision-making. This study uses secondary data from corporate disclosures, industry reports, and academic literature to examine how sustainability is integrated into product design, pricing, and store operations at four global fashion retailers: H M, Zara, Nike, and Primark. H M and Nike prioritize circular design and supply chain transparency; Zara uses localized eco-efficient manufacturing; while Primark uses cost-driven material advancements. Restructured 7Ps framework shows that conventional marketing and promotional activities may connect with ethical labor rules, carbon neutrality goals, and waste reduction techniques, helping retail operations meet environmental restrictions. This research integrates circular economy theory with realistic retail tactics to transform a core marketing model into a sustainability-focused operational framework for the first time. Practical solutions enable fashion shops to reconcile revenue and environmental responsibility while policymakers examine industry-wide sustainable transformations. The study demonstrates that retail adjustments can reduce sectoral footprints by promoting responsible consumption, climate action, and urban sustainability, so directly supporting the UN Sustainable Development Goals (SDG-11, SDG-12, SDG-13).
With increasing urbanization, land contamination caused by improper waste disposal has emerged as a significant environmental challenge. Traditional monitoring methods are often manual, slow, and lack real-time responsiveness. This research proposes an automated system for real-time land contamination detection using computer vision and machine learning, aimed at overcoming the limitations of existing approaches. The system leverages the lightweight MobileNetV2 model to analyze visual data from drones or cameras, accurately identifying and quantifying pollution levels. When waste coverage exceeds a 70
In competitive markets, retailers often use trade credit policies to boost product demand, while wholesalers offer quantity discounts to incentivize larger orders. Trade credit boosts sales but incurs costs from financing accounts receivable and managing collections and bad debt expenses, requiring a careful trade-off. Wholesalers typically seek larger retailer orders, with discounts designed to push purchases beyond the economic order quantity (EOQ). This paper examines wholesaler–retailer dynamics where wholesalers design discount schemes, and retailers set consumer credit policies, with end demand sensitive to both price and credit period. Interactions are modelled under two settings: (i) non-cooperative (Stackelberg game) and (ii) cooperative (joint payoff maximization). When determining optimal inventory, quantity discount, and trade credit strategies for the supply chain, a discounted cash flow (DCF) analysis has been applied to evaluate and recognize the cash flows accurately. Numerical illustrations and analysis of parameter sensitivity on a hypothetical dataset provide important managerial insights into the model. The results demonstrate that a cooperative framework maximizes total supply chain profit.
Pneumonia is a respiratory illness that is caused by fungi, bacteria, and viruses affecting the lungs of human beings. The lack of proper detection of pneumonia has caused a substantial impact on the lives of people owing to its life-threatening nature. In this article, a novel Harmonic Pufferfish Optimization Algorithm enabled Convolutional Neural Network Fusion Deep Kronecker Network (HPOA_CNN–DKN) method is developed for pneumonia detection. Initially, the chest X-ray images are obtained from the dataset and passed through the Medav filter to remove noise. Moreover, the processing is performed by the Fully Convolutional Neural Network (FCN) for the lobe segmentation. The resultant lobe segment images are subjected to data augmentation and processes, like shifting, brightness, resizing, and shearing, to increase the training samples, and multiple features are extracted. The detection of pneumonia is achieved through a CNN–DKN framework that integrates the functionalities of CNN with those of DKN. In addition, the weight of CNN–DKN is tuned optimally by HPOA devised by assimilating Harmonic analysis with the Pufferfish Optimization Algorithm (POA) approach. Moreover, the HPOA_CNN–DKN approach recorded a sensitivity of 0.920, a specificity of 0.920, and an accuracy of 0.914.
The integration of Artificial Intelligence (AI) into Green Supply Chain Management (GSCM) is accelerating as firms seek to enhance decision-making, resource efficiency, and environmental performance. This study conducts a scientometric analysis of 1448 Scopus-indexed articles (1999–2025) using VOSviewer and Bibliometrix to map the evolution, key contributors, and emerging themes within AI-driven GSCM. Three dominant thematic clusters are identified: (1) Circular Economy and Industry 4.0, (2) Blockchain Integration, and (3) Sustainability and Reverse Logistics. Results show rapid growth in machine learning, predictive analytics, blockchain, and multi-objective optimization, reflecting a shift toward digitally enabled sustainability solutions. AI’s role in reverse logistics, waste management, and carbon-emissions reduction is growing, yet challenges remain related to data governance, interoperability, scalability, and organizational readiness. The findings extend theoretical understanding by highlighting the relevance of Dynamic Capabilities, Technology–Organisation–Environment (TOE), and socio-technical transition perspectives in explaining AI-enabled transformations in sustainable supply chains. Practically, the study outlines best practices for AI integration, including forecasting, logistics optimization, blockchain-based traceability, and circular economy performance metrics. Future research should address ethical concerns, regional adoption disparities, and the need for governance frameworks to support responsible and scalable AI deployment in GSCM.
Globally, over 300 million people need humanitarian assistance due to natural and anthropogenic disasters. Ensuring social sustainability in humanitarian logistics remains challenging because of the simultaneous need to ensure equity, inclusivity and efficiency. The inclusion of volunteers in humanitarian logistics delivery, including the out-of-pocket expenditures of socially vulnerable groups as part of the total humanitarian logistics cost, has rarely been considered in the existing literature. This paper contributes to the social sustainability literature by developing an integrated humanitarian relief delivery optimisation model that considers volunteer engagement and out-of-pocket expenditures by socially vulnerable groups under conditions of hyperinflation and supply uncertainty. This study proposes an integer non-linear programming model using a robust optimisation approach to minimise the total relative regret across probabilistic supply scenarios. The model considers the total humanitarian logistics cost (THLC), including government costs for facility operationalisation, procurement, transportation, volunteer incentives, and labour deployment. The model newly considers social sustainability by accommodating the costs borne by socially vulnerable groups in the form of out-of-pocket expenditure incurred to buy essential commodities at hyperinflated rates due to supply insufficiencies. The key decisions include facility operationalisation decision, routing decisions, and labour engagement decisions among others. The model was validated a set of three different scales of problem instances with varying sizes inspired by cyclones in India. A self-tailored particle swarm optimization (s-PSO) approach was employed to solve the proposed model. The results indicated the superiority of s-PSO in comparison to genetic algorithm ( in achieving lower THLC with a reasonable execution time. Sensitivity analysis validated the model’s adaptability to variations in supply, demand and hyperinflation rates.
