A carregar...
Projeto de investigação
Information Management Research Center
Financiador
Autores
Publicações
Understanding technological, cultural, and environmental motivators explaining the adoption of citizen science apps for coastal environment monitoring
Publication . Cardoso-Andrade, Mariana; Cruz-Jesus, Frederico; Souza Troncoso, Jesus; Queiroga, Henrique; Gonçalves, Jorge Manuel Santos
Environmental and nature conservation authorities are calling for a collective effort to break or reduce the current cycle of environmental degradation. Much of the response depends on scientific knowledge production based on thematically and geographically comprehensive datasets. Citizen science (CS) is a cost-effective support tool for scientific research that provides means for building large and comprehensive datasets and promoting public awareness and participation. One of the greatest challenges of CS is to engage citizens and retain par-ticipants in the project. Our work addresses this challenge by (1) defining the role that technological, cultural, and environmental dimensions play in the adoption of CS apps for coastal environment monitoring, and (2) providing base knowledge about the profile of the apps' most likely users and the functional features they require to be successful. Collectivists and people who assume a green identity are the most likely users of these apps. Drivers of their use are the promotion of citizen empowerment, habit development, provision of facilitating conditions, and proof of environmental performance.The outcome of this study is a set of guidelines for project managers, app developers, and policymakers for citizens' engagement and retention in CS coastal environment monitoring projects through their apps.
Large language models powered aspect-based sentiment analysis for enhanced customer insights
Publication . Água, Mariana; António, Nuno; Carrasco, Paulo; RASSAL, CARIMO
In the age of social networks, user-generated content has become vital for organizations in tourism and hospitality. Traditional sentiment analysis methods often struggle to process large volumes of data and capture implicit sentiments. This study examines the potential of Aspect-Based Sentiment Analysis (ABSA) using Large Language Models (LLMs) to enhance sentiment analysis. By employing GPT-4o via ChatGPT, we benchmark three approaches: a fuzzy logic-based method, manual human analysis, and a new ChatGPT-based analysis. We analyze a dataset of 500 all-inclusive hotel reviews, comparing these methods to assess ChatGPT's effectiveness in identifying nuanced language and handling subjectivity. The findings reveal a high similarity between ChatGPT and human analysis, showcasing ChatGPT's ability to interpret complex sentiments and automate sentiment classification tasks. This study highlights the potential of LLMs in transforming customer feedback analysis, providing deeper insights, and improving responsiveness in the hospitality industry. These results contribute to academia by presenting a framework for using LLMs in ABSA and guiding future applications and development.
Improving trust in online reviews: a machine learning approach to detecting artificial intelligence-generated reviews
Publication . Ana Marta Santos; Antonio, Nuno
In the hotel industry, social reputation is critical. Consumers increasingly rely on online reviews for accommodation decisions, making Artificial Intelligence (AI) generated fraudulent reviews a significant threat. Distinguishing between genuine and AI-generated reviews is essential for hotels to maintain credibility. This study creates a unique dataset of AI-generated reviews and combines vectorization methods with text-based features to build a Machine Learning model for identifying nongenuine reviews. Results show that incorporating text-based features significantly improves detection accuracy, and simpler vectorization methods can be effective for simpler datasets. This study contributes to academia by providing a novel methodology and publicly available dataset for further research, and to the hotel industry by enhancing credibility and consumer trust through better review filtering.
Understanding risk factors of post-stroke mortality
Publication . Castro, David; Antonio, Nuno; Marreiros, Ana; Nzwalo, Hipólito
Stroke is one of the leading causes of death worldwide. Understanding the risk factors for poststroke mortality is crucial for improving patient outcomes. This study analyzes and predicts poststroke mortality using the modified Rankin Scale (mRS), a functional neurological evaluation scale. Several Machine Learning models were developed and assessed using a dataset of 332 stroke patients from Hospital de Faro, Portugal, from 2016 to 2018. The Random Forest model outperformed others, achieving an accuracy of 98.5% and a recall of 91.3. Twenty-four risk factors were identified, with stroke severity as the most critical. These findings provide healthcare professionals with valuable tools for early identification and intervention for high-risk stroke patients, enabling informed decision-making and customized treatment plans. This research advances healthcare predictive analytics, offering a precise mortality prediction model and a comprehensive analysis of risk factors, potentially improving clinical outcomes and reducing mortality rates. Future applications could extend to patient monitoring and management across various medical conditions.
Predictive factors driving positive awake test in carotid endarterectomy using machine learning
Publication . Pereira-Macedo, Juliana; Duarte-Gamas, Luís; Pereira Pias, Ana Daniela; Myrcha, Piotr; Andrade, José P.; António, Nuno; Marreiros, Ana; Rocha-Neves, João
Background: Positive neurologic awake testing during the carotid cross-clamping may be present in around 8% of patients undergoing carotid endarterectomy (CEA). The present work aimed to assess the accuracy of an artificial intelligence (AI)-powered risk calculator in predicting intraoperative neurologic deficits (INDs). Methods: Data was collected from carotid interventions performed between January 2012 and January 2023 under regional anesthesia. Patients with IND were selected along with consecutive controls without IND in a case-control study design. A predictive model for IND was developed using machine learning, specifically Extreme Gradient Boosting (XGBoost) model, and its performance was assessed and compared to an existing predictive model. Shapley Additive exPlanations (SHAP) analysis was employed for the model interpretation. Results: Among 216 patients, 108 experienced IND during CEA. The AI-based predictive model achieved a robust area under the curve of 0.82, with an accuracy of 0.75, precision of 0.88, sensitivity of 0.59, and F1Score of 0.71. High body mass index (BMI) increased contralateral carotid stenosis, and a history of limb paresis or plegia were significant IND risk factors. Elevated preoperative platelet and hemoglobin levels were associated with reduced IND risk. Conclusions: This AI model provides precise IND prediction in CEA, enabling tailored interventions for high-risk patients and ultimately improving surgical outcomes. BMI, contralateral stenosis, and selected blood parameters emerged as pivotal predictors, bringing significant advancements to decision-making in CEA procedures. Further validation in larger cohorts is essential for broader clinical implementation.
Unidades organizacionais
Descrição
Palavras-chave
Contribuidores
Financiadores
Entidade financiadora
Fundação para a Ciência e a Tecnologia
Programa de financiamento
6817 - DCRRNI ID
Número da atribuição
UIDB/04152/2020