This paper discusses that the Digital marketplaces play a crucial role in connecting startups with potential investors, yet their adoption success depends on understanding the key factors influencing user intention. Predicting adoption behaviors accurately can help improve engagement and ensure platform sustainability. The study applies the Unified Theory of Acceptance and Use of Technology (UTAUT) framework to identify key adoption factors including Performance Expectancy (PE), Effort Expectancy (EE), Social Influence (SI), Facilitating Conditions (FC), Trust (TR), and Government Support (GS).and this has been widely applied to study technology adoption, limited research integrates this framework with machine learning models to predict adoption intention in e-marketplaces for startups. This study aims to develop machine learning-based prediction models for StartSmart an e-marketplace linking startups and investors and identify the most influential factors affecting adoption intention based on the UTAUT framework. Data from 542 respondents were analyzed using six machine learning techniques: Decision Trees (DT), Random Forests (RF), Gradient Boosting (GRB), XGBoost (XGB), K-Nearest Neighbors (KNN), and Support Vector Machines (SVM).Results indicate that DT achieved the highest accuracy (0.93) and precision (0.94), while RF obtained the highest AUC-ROC score (0.98). Analysis of feature importance revealed that PE and EE were the most significant predictors of adoption, followed by TR and GS. These findings provide valuable insights for platform developers to prioritize usability and performance improvements, and for policymakers to strengthen trust and government support. The study also highlights the potential of combining UTAUT with machine learning to enhance predictive accuracy in digital adoption research.

The accuracy of classification algorithms is significantly influenced by the quality and structure of input data. In this light, effective pre-processing is crucial for boosting the generalization capabilities of supervised machine learning models. This study addresses key challenges in data preparation, including the treatment of continuous attributes, imputation of missing values, and management of high-dimensional features. To overcome these obstacles, we propose an innovative hybrid preprocessing strategy that synthesizes multiple techniques into a unified framework. By tailoring specific methods to the characteristics of diverse datasets, this hybrid approach enhances both the accuracy and robustness of the classification results. Through the promotion of intelligent, data-driven solutions that can be applied in multiple sectors, the findings support the Sustainable Development Goal 9 (SDG 9), which focuses on Industry, Innovation and Infrastructure.

Marine ecosystems face unprecedented threats from pollution, overfishing, and climate change, creating an urgent need for conservation initiatives. While consumer awareness of ocean degradation is increasing, there remains a persistent gap between environmental concern and actual purchasing behavior toward sustainable products. This study aims to examine public readiness to adopt OceanGuardian, a sustainable e-commerce platform for marine conservation products, by integrating behavioral, technological, and environmental perspectives. Using a modified Extended Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) framework, survey data from 600 respondents were analyzed with machine learning models, including Support Vector Machines, Random Forests, and XGBoost, to identify key determinants of consumer intention and use behavior. Results indicate that social influence, performance expectancy, affordability, and habit formation significantly predict adoption, with Support Vector Machines achieving the highest predictive accuracy (92.5%). The findings highlight the potential of artificial intelligence to enhance consumer behavior analysis while recognizing challenges such as economic barriers and consumer skepticism. The study offers theoretical contributions by extending UTAUT2 with environmental factors and provides practical insights for policymakers and businesses to design strategies that foster sustainable shopping and strengthen marine conservation efforts.

Upcycling is a sustainable solution that mitigates environmental changes, focusing on climate action, by extending the lifecycle of products and reducing wastage. This study investigates Filipino consumers’ behavioral intention to use an upcycling e-marketplace, highlighting the intersection of sustainability, consumer psychology, and digital platforms. Despite growing interest in circular economy models, adoption drivers in this domain remain underexplored and are rarely modeled with predictive analytics. To address this gap, the study collected data from 500 Filipino participants capturing environmental knowledge and concern, perceived ease of use and usefulness, attitude toward use, perceived behavioral control, subjective norms, user demand, and intention, then analyzed using multiple machine learning algorithms, namely, Decision Tree, Random Forest, Gradient Boosting, XGBoost, K-Nearest Neighbors, and Support Vector Machine. Among these, the Decision Tree model demonstrated the best balance of predictive accuracy (95%), precision (96%), and recall (91%), suggesting strong classification capability. Analysis of the importance of features revealed that Perceived Usefulness, Attitude Toward Use, and Subjective Norms were the most influential predictors of adoption, outweighing traditional environmental concerns. These findings underscore the importance of designing upcycling platforms that emphasize practical value, user convenience, and social validation. The study concludes that sustainable behavior is more likely when aligned with personal benefits and peer influence, rather than relying solely on environmental appeals.

This study examines the Canvas’ usability in Learning Management System (LMS) from the perspectives of students and teachers, focusing on experiences across Progressive Web App (PWA) and native mobile platforms. A task-based usability testing approach was employed, combining quantitative measures of task completion and time with qualitative insights from observations and participant feedback. Findings indicate that both platforms supported high task completion, though clear differences emerged in efficiency and feature accessibility. Teachers achieved a 91.7% completion rate on the mobile app compared to 100% on the PWA. The mobile app was faster for grading and assignment creation, while the PWA provided broader feature coverage, particularly for analytics, though some users reported navigation difficulties. For students, performance differences were more pronounced: average task completion time on the PWA was 1.24 minutes compared to 5.72 minutes on the mobile app. Tasks such as replying to announcements and checking grades were completed up to ten times faster on the PWA. Overall, the mobile app demonstrated greater stability and efficiency for routine functions, whereas the PWA offered extended functionality and cross-platform access but with tradeoffs in responsiveness and interface clarity. These results highlight the role of platform choice in shaping user experience and suggest directions for optimizing Canvas LMS for both teaching and learning contexts. By advancing usability in digital learning platforms, this research contributes to Sustainable Development Goal (SDG) 4: Quality Education, while also supporting SDG 9: Industry, Innovation, and Infrastructure through insights on mobile technology design, and SDG 10: Reduced Inequalities by emphasizing accessibility across diverse devices and connectivity conditions.

Despite the potential for e-commerce to boost productivity and market access for farmers, adoption remains low, particularly in rural areas of developing countries. This study addresses the research gap by predicting farmers' adoption of digital platforms in the National Capital Region, Philippines, using the Theory of Planned Behavior (TPB) and the Technology Acceptance Model (TAM). Based on a survey of 615 farmers and analysis with various machine learning models, with XGBoost as the top performer, the study found that perceived usefulness, trust, and price value are the most significant factors influencing adoption. Social influence and ease of use also play important roles. The findings provide guidance for policymakers and platform developers, highlighting the need to improve digital literacy, build trust, and ensure affordability to accelerate the digital transformation of the agricultural sector.

High voltage engineering has evolved rapidly, driven by the growing need for efficient energy transmission and the integration of renewable energy into modern power grids, including urban areas. Innovations such as HVDC systems are central to this transformation, ensuring that grids can handle the increasing complexity and demand for sustainable energy. However, challenges remain, especially when it comes to coordinating insulation in hybrid AC/DC systems and maintaining the resilience of the overall infrastructure. This review looks at how Artificial Intelligence (AI) can help tackle these challenges, focusing on its role in fault detection, predictive maintenance, and improving system reliability. By comparing traditional methods with AI-driven solutions, we highlight how AI can enhance the scalability, efficiency, and adaptability of power systems. With AI, utilities can predict and prevent faults, optimize grid performance, and seamlessly integrate renewable energy sources into both rural and urban environments. Our goal is to provide insights for researchers, industry professionals, and policymakers on how AI can be harnessed to build more sustainable, resilient, and reliable energy systems. The insights shared here aim to help shape the future of power grids, positioning AI as a key player in the transition to cleaner, more efficient energy solutions.

The integration of artificial intelligence (AI) into computational materials science (CMS) has introduced powerful approaches for accelerating the discovery and optimization of advanced energy materials. As energy demands shift toward renewable systems, the development of efficient materials for batteries, fuel cells, and electrocatalysts becomes increasingly critical. This paper systematically reviews recent AI methodologies applied within CMS, particularly those leveraging density functional theory (DFT), molecular dynamics (MD), and kinetic Monte Carlo (KMC) simulations. Emphasis is placed on the use of machine learning (ML) models, including supervised learning, deep learning, and hybrid strategies for property prediction, structure optimization, and inverse design. The review categorizes current applications across key energy technologies and discusses how AI is reshaping material screening and development pipelines. It concludes with an outlook on future directions, highlighting the need for standardized datasets, interpretable models, and physics-informed frameworks to improve predictive accuracy and facilitate AI adoption in practical materials research.

Artificial Intelligence (AI) is transforming the prediction and optimization of advanced energy materials by enabling accurate, scalable modeling beyond traditional methods. This review evaluates recent AI applications—including Graph Neural Networks (GNNs), Convolutional and Recurrent Neural Networks (CNNs, RNNs), tree-based ensembles, and Gaussian Process Regression (GPR)—for forecasting performance metrics such as overpotential, conductivity, capacity, and degradation. GNNs achieved R2 > 0.90 in structure-sensitive tasks; LSTM models predicted battery degradation with <10% error; and tree-based models balanced accuracy (MAE < 0.15 V) with interpretability. GPR excelled in low-data regimes via uncertainty quantification. Hybrid and physics-informed models improved generalizability and data efficiency. While challenges remain in data quality and integration with experiments, emerging strategies like autonomous labs and generative design offer promising advances. This review provides comparative benchmarks and highlights pathways for robust AI-driven materials discovery.

The widespread use of Liquefied Petroleum Gas (LPG) in households and industries presents a significant risk due to its highly flammable nature, making early detection of leaks for preventing accidents. Traditional LPG detection methods often rely on manual monitoring, which may not provide timely alerts or automatic responses to prevent accidents. The Internet of Things (IoT) refers to a network of physical objects, or “things,” embedded with sensors, software, and other technologies that enable them to connect and exchange data with other devices and systems over the internet. The advantage of the Internet of Things (IoT) offers new opportunities to safety systems through real-time monitoring and remote alerts. By integrating IoT technology with LPG leak detection, it is possible to create a responsive and reliable safety system. This device focuses on the development of an IoT-based LPG tank leakage detection system with alarm and auto-off system, designed to detect leaks, sound an alarm, and automatically shut off the gas supply or even remotely turn off the valve through an android-based application to prevent accidents. The system uses IoT for real-time data transmission and remote monitoring, allowing users to receive instant notifications and take immediate action, even when away from the premises. The system uses gas sensors to continuously monitor LPG levels, triggering an immediate response in the event of a leak. Integrated with IoT technology, the system provides real-time alerts to users via android-based application, ensuring prompt action.