Artificial intelligence (AI) is gradually being accepted into education and has the potential of transforming learning and teaching. It is therefore important to understand the perceptions of various stakeholders on the impact of AI in engineering education. This study is framed with the Technology Acceptance Model (TAM) and employs a mixed-method approach to assess the perception of engineering students on the adoption of AI. The quantitative analysis of the survey findings using partial least square structural equation modelling (PLS-SEM), reveals that perceived usefulness and the ease of use has a statistically significant relationship with AI adoption amongst engineering students. Moreover, their attitude towards AI did not influence their intention to use AI in their education. Through the qualitative analysis, the perceived usefulness of AI in personalized learning, solving complex problems, increasing efficiency and solving global problems were highlighted. However, it was noted that wider adoption of AI would be achieved when there are strategies to address the institutional barriers, ethical issues and funding. Another interesting finding is that students raised concern about the potential impairment in their cognitive abilities if they over-rely on AI.

Oil Spills continue to pose significant risks across offshore production, marine transport, and subsea infrastructure. Deep Learning (DL) methods now support proactive Oil Spill Prevention (OSP) through Predictive Maintenance, Anomaly Detection, and Visual Surveillance. These approaches often achieve high performance but rely on computationally intensive workflows that demand substantial energy and generate carbon emissions. Such impacts remain largely unquantified in existing OSP research. This paper is a systematic review of DL-based OSP research published between 2020 and mid-2025, reported in accordance with the PRISMA 2020 guidelines. Ninety studies were analyzed to assess AI lifecycle practices across model architectures, data modalities, training regimes, evaluation methods, deployment choices, and sustainability considerations. Results show persistent reliance on accuracy-centric evaluation, with limited reporting of lifecycle energy use, carbon footprint, or hardware-related environmental impacts. To address this gap, a lifecycle-oriented sustainability framework is introduced. The framework operationalizes four metrics: Job Sustainability Cost, Amortized Sustainability Cost, Embodied Product Cost, and Expanded Amortized Sustainability Cost. These metrics capture operational energy demand, carbon intensity, embodied hardware emissions, and reuse effects across the AI lifecycle. Thus, supporting model comparison, compression strategies, carbon-aware scheduling, and lightweight performance logging. They integrate into routine OSP workflows by offline profiling of energy usage and carbon emissions during model development, evaluation, and scheduled audits. Profiling is performed outside live control loops, ensuring that safety-critical operational logic remains unaffected. By combining DL methodologies with lifecycle-based assessment, this study demonstrates how accuracy-centric evaluation can be complemented by transparent, resource-aware analysis. The proposed framework applies to OSP and is transferable to other safety-critical industrial monitoring domains.

This bibliometric review analyses 6927 studies indexed in Scopus and Web of Science between 2014 and 2025 to trace the evolution of human-centric digital transformation in digital business environments. Using performance and science-mapping techniques, this study identifies strong publication growth and four major research streams: technology adoption models, behavioural and well-being dimensions, capability theories, and Industry 4.0–5.0 frameworks. The findings indicate a progression from technology-led digital transformation towards capability building and, more recently, human-centric approaches that emphasise digital skills, resilience, leadership, sustainability, and ethical human–machine collaboration. The results provide an integrated overview of the intellectual structure of the field, and show how human-centric perspectives are becoming increasingly important in digitally mediated organisational settings. The study concludes by outlining a future research agenda for human-centric digital transformation in digital business environments.

Among other constraints, limited energy availability is a major challenge in Wireless Sensor Networks (WSNs). Therefore, conserving energy is crucial for enhancing the efficiency and prolonging the lifetime of WSNs. The hierarchical model, particularly clustering, has become an effective mechanism for achieving optimal energy utilisation. Clustering in a WSN helps to organise nodes into clusters, each with at least one leader. However, introducing cluster leaders (CLs) into WSNs can negatively impact packet delivery, and the CL selection algorithm may lead to network instability. The paper proposes a more efficient CL selection technique, building on VERD, an existing algorithm that leverages the sensors' Visibility index, Energy, Received Signal Strength Indicators, and Euclidean Distance from the base station as parameters for the CL eligibility index. The Improved VERD (IVERD), proposed in this work, identifies the optimal CL selection in WSNs by using the Grey Wolf Optimiser to determine the relative coefficient values of the CL eligibility parameters, leading to a substantial enhancement in WSN energy management and sensor lifetime. In comparative analysis with algorithms like VERD, HEEL (coined from the selection parameters - Hop counts, residual Energy, neighbour Energy and number of Links), Stable Election Protocol (SEP), and Power-Efficient Gathering in Sensor Information Systems (PEGASIS), IVERD demonstrates a significant improvement in WSN energy conservation, extending sensor lifetime by 36.94%. It also enhanced network stability and packet delivery by 34.11% and 24.46%, respectively. Our research shows that IVERD offers a more sustainable and robust solution for real-world applications of WSNs.

The proliferation of user-generated content in today’s digital landscape has further increased dependence on online reviews as a source for decision-making in the hospitality industry. There has been an increasing interest in automating this decision-support mechanism through recommender systems. However, this process often requires a large amount of labelled corpus to train an effective algorithm, necessitating the use of human annotators for developing training data, where this is lacking. Although the manual annotation can be helpful in enriching the training corpus, it can, on the one hand, introduce errors and annotator bias, including subjectivity and cultural bias, which can affect the quality of the data and fairness in the model. This paper examines the alignment of ratings derived from different annotation sources and the original ratings provided by customers, which are treated as the ground truth. The paper compares the predictions from Generative Pre-trained Transformer (GPT) models against ratings assigned by Amazon Mechanical Turk (MTurk) workers. The GPT 4o annotation outputs closely mirror the original ratings, given its strong positive correlation (0.703) with the latter. The GPT-3.5 Turbo and MTurk showed weaker correlations (0.663 and 0.15, respectively) than GPT 4o. The potential cause of the large difference between original ratings and MTurk (largely driven by human perception) lies in the inherent challenges of subjectivity, quantitative bias, and variability in context comprehension. These findings suggest that the use of advanced models such as GPT-4o can significantly reduce the potential bias and variability introduced by Amazon MTurk annotators, thus improving the prediction accuracy of ratings with actual user sentiment as expressed in textual reviews. Moreover, with the per-annotation cost of an LLM shown to be thirty times cheaper than MTurk, our proposed LLM-based textual review annotation approach will be cost-effective for the hospitality industry.

Humans face various diseases that are mainly caused by environmental conditions and living habits. These diseases exhibit several symptoms and can share a relationship based on their symptoms. The identification and interpretation of these groups of symptom-based diseases can aid in developing treatment plans for a new outbreak of disease. This research explores the intersection of machine learning and healthcare, specifically focusing on the enhancement of disease classification through symptom-based cluster analysis. By leveraging unsupervised machine learning algorithms, patterns and relationships within diverse symptom datasets were identified, revealing novel associations and subtypes in disease manifestation. The integration of a Large Language Model (LLM), specifically OpenAI’s Generative Pretrained Transformer(GPT), played a pivotal role in interpreting and communicating the complex outputs of the machine learning process. The results indicated a significant improvement in defining distinct clusters based on the relationship between diseases and symptoms, with GPT-4o providing simplified explanations that bridge the gap between machine-generated insights and healthcare professional’s understanding. The study’s findings offer a more profound understanding of the distinctive features characterising the different clusters of diseases generated by the machine learning models.

There has been a growing environmental concern related to artificial intelligence (AI) and the need to optimize AI models for greater energy efficiency and reduced carbon emissions. This study analysed the key contributors, institutions, countries, journals, collaborations and current developments in this field. By analyzing 385 articles published between 2016 and 2024, retrieved from the Web of Science database, and utilizing tools like Microsoft Excel and VOSviewer, the study provides a global review of Green AI research, with a particular focus on Africa. The findings reveal a rise in Green AI research since 2020, with the USA leading in research output and international collaboration, while African countries ranked low. Key themes include energy efficiency, carbon footprint reduction, and the development of sustainable AI models.

Feature selection has become an important step in machine learning pipelines, contributing to model interpretability and accuracy. While the emphasis has been hugely on global feature selection techniques, these methods do not support feature attributions to the distinct groups within a dataset, since they assume that a single feature set is adequate to correctly undertake the classification task. Unlike unsupervised learning, moreover, feature selection techniques, whether global or local, have been well-developed for supervised learning. Due to the preceding reasons, this paper presents ClusterSwarm, a new approach towards cluster-based feature selection using Binary Particle Swarm Optimisation (BPSO) and the K-means algorithm, to identify cluster-specific feature sets. Evaluating using four publicly available datasets from the UCI repository, ClusterSwarm demonstrates superior performance to the standard K-means algorithm and agglomerative hierarchical clustering and performs similarly to Sparse K-means, a global feature selection technique. However, ClusterSwarm performs better than Sparse K-means in high-dimensional, multi-class and noisy contexts, while providing interpretability through feature attributions to each cluster. In comparison with CS Sparse K-means, a cluster-specific variant of Sparse K-means, ClusterSwarm produced better accuracies and more efficient feature selection, ignoring redundant features, unlike CS Sparse K-means. In addition to the four public datasets, we experimented with two synthetic datasets carefully curated to represent cases of noisy features and overlapping clusters. These datasets have been used to demonstrate the superiority of ClusterSwarm compared to Sparse K-means, CS Sparse K-means and the standard clustering techniques.

BackgroundThis systematic literature review examines the evolution of digital transformation in e-commerce from 2017 to 2024, with a focus on the intersection of technology, organisational priorities, and human-centric insights aligned with Industry 5.0. The review addresses two key questions: (1) What are the key characteristics and trajectory of research on e-commerce digital transformation? and (2) How have the technological, organisational, and human-centric themes evolved over time?MethodWe searched Scopus, Web of Science, and IEEE Xplore for articles published up to June 30, 2024. Screening was conducted in accordance with PRISMA 2020 guidelines, using predefined inclusion and exclusion criteria. A total of 119 studies were included after a four-stage screening process. Descriptive statistics and thematic synthesis were used for data analysis.ResultsThe results show a steady growth trend on the topic, with 119 studies published across 55 countries and 96 journals, with China as the main contributor. Research evolved through three phases: Early Years (2017–2019): the foundational exploration era, Increased Momentum (2020–2021): strategic adaptation era, and Growth (2022–2023): resilience, personalisation, and sustainability era.ConclusionKey technologies identified include AI, big data, CRM, blockchain, and IoT, with increasing emphasis on ethical and human-centric innovation. Findings highlight a paradigm shift toward human-centric digital transformation, predicated on the ethical use of technology, agile leadership, and data-driven personalisation. Limitations include excluding conference papers and restricted access to some full texts of relevant articles.

The rig state plays a crucial role in recognizing the operations carried out by the drilling crew and quantifying Invisible Lost Time (ILT). This lost time, often challenging to assess and report manually in daily reports, results in delays to the scheduled timeline. In this paper, the Naive Bayes algorithm was used to establish a novel rig state. Training data, consisting of a large set of rules, was generated based on drilling experts’ recommendations. This dataset was then employed to build a Naive Bayes classifier capable of emulating the cognitive processes of skilled drilling engineers and accurately recognizing the actual drilling operation from surface data. The developed model was used to process high-frequency drilling data collected from three wells, aiming to derive the Key Performance Indicators (KPIs) related to each drilling crew’s efficiency and quantify the ILT during the drilling connections. The obtained results revealed that the established rig state excelled in automatically recognizing drilling operations, achieving a high success rate of 99.747

Stroke poses a significant global health challenge, contributing to widespread mortality and disability. Identifying predictors of stroke risk is crucial for enabling timely interventions, thereby reducing the increasing impact of strokes. This research addresses this imperative by employing Explainable Artificial Intelligence (XAI) techniques to pinpoint stroke risk predictors. To bridge existing gaps, six machine learning models were assessed using key performance metrics. Utilising the Synthetic Minority Over-sampling Technique (SMOTE) to minimize the impact of the imbalanced nature of the dataset used in this research, the Random Forest algorithm emerged as the most effective among the algorithms with an accuracy of 94.5%, AUC-ROC of 0.95, recall of 0.96, precision of 0.93, and an F1 score of 0.95. This study explored the interpretation of these algorithms and results using Local Interpretable Model-agnostic Explanations (LIME) and Explain Like I’m Five (ELI5). With the interpretations, healthcare providers can gain insight into patients’ stroke risk predictions. Key stroke risk factors highlighted by the study include Age, Marital Status, Glucose Level, Body Mass Index, Work Type, Heart Disease, and Gender. This research significantly contributes to healthcare and healthcare informatics by providing insights that can enhance strategies for stroke prevention and management, ultimately leading to improved patient care. The identified predictors offer valuable information for healthcare professionals to develop targeted interventions, fostering a proactive approach to mitigating the impact of strokes on individuals and the healthcare system.

Across the globe, governments are developing policies and strategies to reduce carbon emissions to address climate change. Monitoring the impact of governments’ carbon reduction policies can significantly enhance our ability to combat climate change and meet emissions reduction targets. One promising area in this regard is the role of artificial intelligence (AI) in carbon reduction policy and strategy monitoring. While researchers have explored applications of AI on data from various sources, including sensors, satellites, and social media, to identify areas for carbon emissions reduction, AI applications in tracking the effect of governments’ carbon reduction plans have been limited. This study presents an AI framework based on long short-term memory (LSTM) and statistical process control (SPC) for the monitoring of variations in carbon emissions, using UK annual CO2 emission (per capita) data, covering a period between 1750 and 2021. This paper used LSTM to develop a surrogate model for the UK’s carbon emissions characteristics and behaviours. As observed in our experiments, LSTM has better predictive abilities than ARIMA, Exponential Smoothing and feedforward artificial neural networks (ANN) in predicting CO2 emissions on a yearly prediction horizon. Using the deviation of the recorded emission data from the surrogate process, the variations and trends in these behaviours are then analysed using SPC, specifically Shewhart individual/moving range control charts. The result shows several assignable variations between the mid-1990s and 2021, which correlate with some notable UK government commitments to lower carbon emissions within this period. The framework presented in this paper can help identify periods of significant deviations from a country’s normal CO2 emissions, which can potentially result from the government’s carbon reduction policies or activities that can alter the amount of CO2 emissions.

The healthcare sector has suffered from wastage of resources and poor service delivery due to the significant impact of appointment no-shows. To address this issue, this paper uses explainable artificial intelligence (XAI) to identify major predictors of no-show behaviours among patients. Six machine learning models were developed and evaluated on this task using Area Under the Precision-Recall Curve (AUC-PR) and F1-score as metrics. Our experiment demonstrates that Support Vector Classifier and Multilayer Perceptron perform the best, with both scoring the same AUC-PR of 0.56, but different F1-scores of 0.91 and 0.92, respectively. We analysed the interpretability of the models using Local Interpretable Model-agnostic Explanation (LIME) and SHapley Additive exPlanations (SHAP). The outcome of the analyses demonstrates that predictors such as the patients' history of missed appointments, the waiting time from scheduling time to the appointments, patients' age, and existing medical conditions such as diabetes and hypertension are essential flags for no-show behaviours. Following the insights gained from the analyses, this paper recommends interventions for addressing the issue of medical appointment no-shows.

Governments have implemented a variety of national and international efforts to reduce carbon emissions to prevent the damaging effects of climate change on the environment and the global economy through the execution of several policies, including the Paris Agreement. Achieving the objectives of the Paris Agreement seeks to keep the increase in average global temperature to well below 2 ℃ and, preferably, below 1.5 ℃ above pre-industrial levels, will need a shift towards renewable energy sources like solar and wind power. As a result of these efforts, renewable energy sources’ capacity is projected to expand in the upcoming years. Offshore wind is the UK's leading renewable energy source for power generation. Recently, cost optimisation efforts in the offshore wind industry have been through engineering design, especially with increasing turbine capacities. This research demonstrates how data analytics can achieve cost optimisation in procurement activities of offshore wind projects, which presents opportunities to reduce the Levelized Cost of Energy (LCOE). Ten (10) offshore wind projects have been selected in the UK offshore wind industry by building a workflow and designing an analytic app using Alteryx Designer. The workflow is built on procurement quotations of wind turbines, marine vessels, and export cables to optimise procurement costs and delivery timelines while fulfilling the established constraints for the respective projects. The optimisation results identified three areas of cost-optimisation opportunities: Contracting, Collaboration, and Reusing Components.

State-of-the-art autonomous AI algorithms such as reinforcement learning and deep learning techniques suffer from high computational complexity, poor explainability ability, and a limited capacity for incremental adaptive learning. In response to these challenges, this paper highlights the TMGWR-based algorithm, developed by the present authors, as a case study towards self-adaptive unsupervised learning in autonomous developmental AI, and makes the following contributions: it presents and reviews essential requirements for today’s autonomous AI and includes analysis for their potential for Green AI; it demonstrates that, unlike these state-of-the-art algorithms, TMGWR possesses explainability potentials that can be further developed and exploited for autonomous learning applications. In addition to shaping researchers’ choice of metrics for selecting autonomous learning strategies, this paper will help to motivate further innovative research in autonomous AI.

This article presents a novel Artificial Intelligence (AI) workflow to enhance drilling performance by mitigating the adverse impact of drill-string vibrations on drilling efficiency. The study employs three supervised machine learning (ML) algorithms, namely the Multi-Layer Perceptron (MLP), Support Vector Regression (SVR), and Regression Decision Tree (DTR), to train models for bit rotation (Bit RPM), rate of penetration (ROP), and torque. These models combine to form a digital twin for a drilling system and are validated through extensive cross-validation procedures against actual drilling parameters using field data. The combined SVR - Bit RPM model is then used to categorize torsional vibrations and constrain optimized parameter selection using the Particle Swarm Optimization block (PSO). The SVR-ROP model is integrated with a PSO under two constraints: Stick Slip Index (SSI<0.05) and Depth of Cut (DOC<5 mm) to further improve torsional stability. Simulations predict a 43% increase in ROP and torsional stability on average when the optimized parameters WOB and RPM are applied. This would avoid the need to trip in/out to change the bit, and the drilling time can be reduced from 66 to 31 h. The findings of this study illustrate the system's competency in determining optimal drilling parameters and boosting drilling efficiency. Integrating AI techniques offers valuable insights and practical solutions for drilling optimization, particularly in terms of saving drilling time and improving the ROP, which increases potential savings.

With the recent global surge in Mpox (formerly Monkeypox) cases, researchers have proposed deep learning technologies for early detection of the disease from skin lesion images. However, many of these researchers follow the current Red AI trend of seeking to improve the performance accuracies of classifiers with no consideration given to the efficiency and environmental-friendliness of their models. This paper proposes a Green AI model selection strategy based on a multi-criteria decision technique, incorporating computational time in identifying the optimal model for final deployment. We have experimented with end-to-end ResNet50, VGG19, and InceptionV3 networks and the transfer-learning of their pre-trained versions with SVMs. Using our proposed Green AI strategy, we have identified the optimal models based on efficiency and performance. The results have been assessed using expert-level validation. We demonstrate that our proposed method can select the best model. The outcomes of our model selection strategy are similar to experts' choices of the optimal model when presented with both model error and computation time. This paper's contributions are significant as they support the ongoing call for Green AI, especially within the healthcare sector.

Well logging has been an integral part of decision making at different stages (drilling, completion, production, abandonment) of a well's history. However, the traditional human-reliant approach to well-log interpretation, which has been the most common practice in the industry, can be time consuming, subjective, and incapable of identifying fine details in log curves. Previous studies have recommended automated approaches as a candidate for addressing these challenges. Despite the progress made so far, what is not yet clear from the existing literature is the extent to which these automated approaches can dispense with human interventions in real -life scenarios. This paper presents an empirical review of different depth-matching techniques in real-life timelapse well logs, primarily focusing on gamma ray and the extent to which the outcomes of these techniques match the results from a human expert. Specifically, the performances of dynamic time warping (DTW), constrained DTW (CDTW), and correlation optimized warping (COW) are investigated. The experiments also consider the effects of filtering and normalization on the performance of each of the techniques. Concerning the correlations of each technique's outcome with the reference data and an expert -generated outcome, this research identifies and discusses its key challenges, as well as provides recommendations for future research directions. Although the COW technique has its limitations, as discussed in this paper, our experiments demonstrate that it shows more potential than DTW and its variants in the well-log pattern alignment task. The work entailed by this research is significant because identifying and discussing the limitations of these techniques is vital for solution-oriented future research in this area.

By harnessing both implicit and explicit customer data, companies can develop a more comprehensive understanding of their consumers, leading to better customer engagement and experience, and improved loyalty.As a result, businesses have embraced many AI technologies, including chatbots, sentiment analysis, voice assistants, predictive analytics, and natural language processing, within customer services and e-commerce.The arrival of ChatGPT, a state-of-the-art deep learning model trained with general knowledge in mind, has brought about a paradigm shift in how companies approach AI applications.However, given that most business problems are bespoke and require specialised domain expertise, ChatGPT needs to be aligned with the requisite task-oriented ability to solve these issues.This paper presents an iterative procedure that incorporates expert system development process models and prompt engineering, in the design of descriptive knowledge and few-shot prompts, as are necessary for ChatGPT-powered expert systems applications within customer services.Furthermore, this paper explores potential application areas for ChatGPT-powered expert systems in customer services, presenting opportunities for their effective utilisation in the business sector.

Due to their dependence on a task-specific reward function, reinforcement learning agents are ineffective at responding to a dynamic goal or environment. This paper seeks to overcome this limitation of traditional reinforcement learning through a task-agnostic, self-organising autonomous agent framework. The proposed algorithm is a hybrid of TMGWR for self-adaptive learning of sensorimotor maps and value iteration for goal-directed planning. TMGWR has been previously demonstrated to overcome the problems associated with competing sensorimotor techniques such SOM, GNG, and GWR; these problems include: difficulty in setting a suitable number of neurons for a task, inflexibility, the inability to cope with non-markovian environments, challenges with noise, and inappropriate representation of sensory observations and actions together. However, the binary sensorimotor-link implementation in the original TMGWR enables catastrophic forgetting when the agent experiences changes in the task and it is therefore not suitable for self-adaptive learning. A new sensorimotor-link update rule is presented in this paper to enable the adaptation of the sensorimotor map to new experiences. This paper has demonstrated that the TMGWR-based algorithm has better sample efficiency than model-free reinforcement learning and better self-adaptivity than both the model-free and the traditional model-based reinforcement learning algorithms. Moreover, the algorithm has been demonstrated to give the lowest overall computational cost when compared to traditional reinforcement learning algorithms.