6277 research outputs found
Sort by
A KPCA–FAHP-based risk analysis of ship crane mishandling
No full textSudden mechanical failure in a ship crane due to mishandling presents significant safety risks for seafarers and dock workers. While previous research has focused on crane maintenance, the risks associated with mishandling during operations remain underexplored. This study introduces a methodology integrating kernel principal component analysis, -distributed stochastic neighbour embedding and tree analysis. A case study examines 10 years of alarm logs from two 35-ton ship cranes, revealing distinct risk patterns. The crane near the bow recorded 51 alarms when lifting loads above the ship, whereas the crane near crew accommodation logged only 22. Alarm frequency varied by location, with 71.3% of all alarms occurring in a single port over 3.8 days. Additionally, operational timing influenced the alarm distribution - alarms peaked between 00:00 and 05:00 near crew accommodations, whereas bow crane alarms were more frequent from 12:00 to 16:00. These findings highlight the need for targeted operational safety measures
Impact of AI-generated phishing attacks: a new cybersecurity threat
No full textAs generative artificial intelligence (AI) technologies, such as OpenAI’s GPT-4 and other large language models (LLMs), advance, the cybersecurity landscape faces a pressing challenge: the rise of AI-generated phishing attacks. These attacks exploit AI’s ability to convincingly mimic human language at scale, making them significantly harder to detect than traditional, human-generated phishing attempts. The purpose of this study is to address this growing threat by analysing the unique characteristics of AI-driven phishing emails, focusing on the linguistic and contextual features that distinguish them from human-authored emails. The research identifies a critical limitation in current phishing detection systems, which exhibit low precision and recall when dealing with AI-generated content. To solve this issue, we propose a novel detection framework that combines advanced machine learning and natural language processing (NLP) techniques with behavioural analysis. Experimental results demonstrate the effectiveness of the proposed model, achieving a precision rate of 91% and a recall rate of 89% and an F1 score of 90% in identifying AI-generated phishing emails—substantially outperforming existing systems. This paper’s main contributions include a comprehensive analysis of AI-generated phishing email characteristics, the development of an innovative detection framework, and the presentation of adaptive strategies to enhance cybersecurity defences against the growing threat of generative AI
Context matters: weaknesses in port State control enforcement of work/rest hours regulations
No full textThe decades of research showing the prevalence of seafarers’ non-compliance with work/rest hours regulations suggest ineffective compliance monitoring and enforcement. This paper examines the practice of inspections by port State control (PSC) officers through 55 semi-structured interviews. Findings show that initial inspections remain simple document consultation. Cross-checking records accuracy is scarce, allowing many instances of non-compliance to go unnoticed. Additionally, PSC officers reported that seafarers skilfully align records, complicating the detection of inconsistencies. Current guideline limitations, time and resource constraints, and the pressure on PSC officers drive them towards prioritising technical issues over human factors-related issues such as fatigue and work/rest hours. Consequently, considering the impacts of PSC officers’ working context, enhancing initial inspections, implementing tamper-proof recording systems, strengthening inspection teams, and updating guidelines are possible options to enforce work/rest hours regulations
HSE-GNN-CP: spatiotemporal teleconnection modeling and conformalized uncertainty quantification for global crop yield forecasting
No full textGlobal food security faces escalating threats from climate variability and resource constraints. Accurate crop yield forecasting is essential; however, existing methods frequently overlook complex spatial dependencies driven by climate teleconnections, such as the ENSO, and lacks rigorous uncertainty quantification. This paper presents HSE-GNN-CP, a novel framework integrating heterogeneous stacked ensembles, graph neural networks (GNNs), and conformal prediction (CP). Domain-specific features are engineered, including growing degree days and climate suitability scores, and explicitly model spatial patterns via rainfall correlation graphs. The ensemble combines random forest and gradient boosting learners with bootstrap aggregation, while GNNs encode inter-regional climate dependencies. Conformalized quantile regression ensures statistically valid prediction intervals. Evaluated on a global dataset spanning 15 countries and six major crops from 1990 to 2023, the framework achieves an R2 of 0.9594 and an RMSE of 4882 hg/ha. Crucially, it delivers calibrated 80% prediction intervals with 80.72% empirical coverage, significantly outperforming uncalibrated baselines at 40.03%. SHAP analysis identifies crop type and rainfall as dominant predictors, while the integrated drought classifier achieves perfect accuracy. These contributions advance agricultural AI by merging robust ensemble learning with explicit teleconnection modeling and trustworthy uncertainty quantification
What an online tea break taught us about collaboration
No full textThe organic, informal daily tea break enjoyed by Solent University’s Learning Design team allowed for serendipitous conversations and the informal exchange of information and practice across networks. Through its post-lockdown transformation into a free-flowing, relatively unstructured team meeting focused on problem-solving, it became a liminal space where our individual networks could meet and cross-pollinate, creating connections and collaborations across the whole university community. Our person-centred approach allows for ownership and empowerment within a context of mutual support and shared insights along with personal authenticity. Together, we are able to create a learning environment that redefines productivity to value people and process over output
A Freirean understanding of the Venezuelan crisis in Trinidad and Tobago
No full textThe Venezuelan crisis, which has seen around 50,000 Venezuelan migrants flee to Trinidad and Tobago, is a continuing humanitarian crisis. While the Venezuelan crisis and its impacts on Venezuela has received considerable scholarly attention, little importance has been giving to the crisis impacts on Venezuela’s neighbours of Trinidad and Tobago (T&T) and the displaced Venezuelans that now reside there. Therefore, this paper explores the demographic change to T&T through the investigation of two separate community initiatives, using Paolo Freire’s ideas of critical consciousness as a theoretical lens. Through a series of semi-structured interviews, the paper examines how the community initiatives have acted as a humanitarian buffer to support the Venezuelan migrants, to the displeasure of some native Afro-centric individuals across Trinidad and Tobago. More specifically, this paper deconstructs the Venezuelan crisis, which was initially lauded as a socialist revolution but quickly became a catastrophic emergency both nationally and internationally, thereby leaving the neighbouring sovereign states to pick up the pieces
Undergraduate students’ perceptions of generative artificial intelligence as a predictor of learning autonomy in Ghana
No full textThe rapid adoption of generative artificial intelligence (GenAI) in higher education raises important questions about students’ preparedness to use these tools meaningfully, particularly in low-resource contexts where digital access and AI literacy remain limited. This study examined whether undergraduates’ perceptions of GenAI use predict learning autonomy in a Ghanaian higher education setting (N = 969). A cross-sectional survey revealed that more than half of the students reported no prior experience with AI tools, and nearly one out of three had low AI literacy levels. Grounded in Self-Determination Theory (SDT), the data analysis applied descriptive statistics, Pearson’s correlation, and linear regression. Findings demonstrated a strong positive relationship between GenAI perception and learning autonomy, with GenAI perception accounting for 75.8% of the variance in autonomy. While this effect size is unusually high and should be interpreted with caution, the study provides actionable insights for designing educational strategies that equip students with the skills to engage responsibly and effectively with AI tools, thereby fostering autonomy and preparing them for a digitally evolving academic environment
Exploring the integration, configuration, and implication of the design process of played-form practice in soccer based on ecological psychology
No full textPlayed-form practice designates training activities for player development or performance outcomes. To achieve targeted outcomes through practice, these activities must be carefully planned so they provide the relevant opportunities for actions together with the right challenges for improvement. In that sense, designing played-form activities that will have intended effects on the performance of participants implies informing key elements of interaction between task, environment, and participants within the settings of the activities. The aim of this paper is to explore the integration of these elements as contributors to the development of players, each one analyzed with respect to psychological theories. Key information about these elements is described along the narrative, with particular attention towards (1) demands, conceived as the feed for task design, as well as (2) the notion of intention, interpreted as a catalyzer of objectivity towards outcomes. The use of a model is expected to guide the design of customized practice settings for team sports like soccer, always with the care to adjust it to the different levels of competition. Implications highlight the contribution of a design model to compensate the current abundance of exemplified suggestions in the practical literature
Ensemble deep learning architectures for detecting pulmonary tuberculosis in chest X-rays
No full textTuberculosis (TB) remains a major global health challenge, causing approximately 1.4 million deaths annually. In many high-burden regions, limited access to expert radiological interpretation leads to delayed or missed diagnoses. To address this, we propose a cost-effective, automated TB screening method suitable for under-resourced settings. Our method integrates a Convolutional Autoencoder Neural Network and a Multi-Scale Convolutional Neural Network with deep layer aggregation into an ensemble learning architecture for robust TB detection from chest radiographs. The framework was evaluated on two public datasets and one private dataset, achieving 99% sensitivity and 94% specificity on the Shenzhen dataset, and consistently high accuracy across all datasets. Expert radiologists reviewed a subset of the predictions, confirming the clinical relevance and diagnostic reliability of the model. The ensemble approach demonstrated strong generalisability, effectively identifying active pulmonary TB in chest X-rays from a globally representative cohort. It also outperformed existing classifiers, achieving a state-of-the-art Area Under the Receiver Operating Characteristic of 0.98. These results highlight the potential of our approach as a practical and scalable tool for TB screening, particularly in low- and middle-income countries where radiological resources are limited
A hybrid framework for assessing near miss reporting culture in Greek ship management
No full textThis study develops a hybrid risk analysis method combined with the FAHP-TOPSIS ranking method for ship management companies to evaluate their performance in occupational risk prevention. Emphasis is given to ships’ near-miss reports weighted with their occupational risks. The proposed ranking system avoids biases favouring specific vessel types or large companies. The proposed methodology ranking system avoids biases favouring larger fleets or specific vessel types. The data was collected from 14 Greek ship management companies managing 167 ships. Initially, the risk analysis revealed that larger companies collect extensive data. However, fleet size and type do not significantly influence reporting trends. Findings highlight that near-miss reporting involves occupational risks related to personal protective equipment, safe movement (including embarkation), health, and work. However, significant underreporting persists in security, pollutant handling, navigation and engine room operations. Following risk analysis, the FAHP-TOPSIS was used to evaluate each company based on the types of near-miss and more frequent reporting that contribute to occupational risk prevention due to weight. The findings show that companies’ preventive culture, as shown in near-miss reporting, is not dependent on the number or characteristics of their fleets. Future research should examine cultural variations in reporting practices beyond the Greek maritime industry to enhance global maritime safety
Data sources:
Access Repository Dashboard
Do you manage Solent University Research Portal? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!