Ismail El Sayad, Felicia Hui Ling Chong, Bhupinder Gosal
Motivated by the increasing volatility in the Canadian housing market, rising interest rates, and the tightening regulatory landscape, this study explores the application and comparative analysis of artificial intelligence (AI) models for assessing mortgage risk, in particular, default risk prediction. Traditional techniques, such as credit scoring and financial ratios, often fail to capture the intricate, non-linear relationships and shifting borrower behaviors characteristic of modern mortgage portfolios. AI-driven models, on the other hand, excel at processing complex datasets and uncovering hidden patterns. This research evaluates multiple AI approaches to assess their predictive accuracy, adaptability, and interpretability, including Artificial Neural Networks (ANNs), Convolutional Neural Networks (CNNs), Random Forests (RF), and XGBoost. Results demonstrate that ensemble models, particularly the class-weighted XGBoost model, deliver superior adaptability to volatility ( i.e ., volatile housing market), while neural networks show potential when applied to rich datasets but demand significant computational resources. By improving borrower risk prediction and enabling proactive adjustments to loan terms, AI models help financial institutions reduce default rates, achieve regulatory compliance, and optimize operational costs. These models enhance the financial sector’s resilience to market volatility while paving a way towards more sustainable lending practices. This study highlights the need to balance predictive performance, interpretability, and adaptability, offering insights for leveraging AI in effective, data-driven mortgage risk management.
This study proposes an advanced computer-aided diagnosis (CAD) framework for thyroid disease diagnosis that integrates numerical patient data and ultrasound images. The framework uses cutting edge technologies, including Vision Transformers (ViTs) and SHapley Additive exPlanations (SHAPs), to increase diagnostic accuracy, interpretability, and clinical applicability. The proposed CAD framework employs the sparse search algorithm (SSA) for optimized feature selection from numerical data and the tree-structured Parzen estimator for tuning the hyperparameters. ViTs are utilized for analyzing thyroid ultrasound images, whereas SHAP provides explainable AI insights into model predictions. Extensive experiments were conducted on two datasets: the thyroid disease patient dataset and the DDTI: Thyroid Ultrasound Images dataset. Performance was evaluated via five-fold and ten-fold cross-validation utilizing metrics including accuracy, precision, and recall. The framework achieved promising performance, with models trained without data augmentation consistently outperforming their augmented counterparts. For the thyroid disease patient dataset, the best-performing model reported an accuracy of 99.71%, precision of 97.05%, recall of 99.29%, and F1-score of 98.16%. For the DDTI dataset, ViTs achieved an accuracy of 95.06% without augmentation, surpassing existing methodologies. Key features such as thyroxine, thyroid surgery, and thyroid-stimulating hormone (TSH) were identified as critical predictors of thyroid conditions. This study underscores the potentiality of AI-driven approaches in healthcare, paving the way for improved diagnostic outcomes and personalized treatment strategies.
The history of Artificial Intelligence (AI) is a narrative of waves -- rising optimism followed by crashing disappointments. AI winters, such as the early 2000s, are often remembered as barren periods of innovation. This paper argues that such a perspective overlooks a crucial wave of AI that seems to be forgotten: the rise of the Semantic Web, which is based on knowledge representation, logic, and reasoning, and its interplay with intelligent Software Agents. Fast forward to today, and ChatGPT has reignited AI enthusiasm, built on deep learning and advanced neural models. However, before Large Language Models (LLMs) dominated the conversation, another ambitious vision emerged -- one where AI-driven Software Agents autonomously served Web users based on a structured, machine-interpretable Web. The Semantic Web aimed to transform the World Wide Web into an ecosystem where AI could reason, understand, and act. Between 2000 and 2010, this vision sparked a significant research boom, only to fade into obscurity as AI's mainstream narrative shifted elsewhere. Today, as LLMs edge toward autonomous execution, we revisit this overlooked wave. By analyzing its academic impact through bibliometric data, we highlight the Semantic Web's role in AI history and its untapped potential for modern Software Agent development. Recognizing this forgotten chapter not only deepens our understanding of AI's cyclical evolution but also offers key insights for integrating emerging technologies.
Andrei Dumitriu, Florin Miron, Florin Tatui
et al.
This report presents an overview of the AIM 2025 RipSeg Challenge, a competition designed to advance techniques for automatic rip current segmentation in still images. Rip currents are dangerous, fast-moving flows that pose a major risk to beach safety worldwide, making accurate visual detection an important and underexplored research task. The challenge builds on RipVIS, the largest available rip current dataset, and focuses on single-class instance segmentation, where precise delineation is critical to fully capture the extent of rip currents. The dataset spans diverse locations, rip current types, and camera orientations, providing a realistic and challenging benchmark. In total, $75$ participants registered for this first edition, resulting in $5$ valid test submissions. Teams were evaluated on a composite score combining $F_1$, $F_2$, $AP_{50}$, and $AP_{[50:95]}$, ensuring robust and application-relevant rankings. The top-performing methods leveraged deep learning architectures, domain adaptation techniques, pretrained models, and domain generalization strategies to improve performance under diverse conditions. This report outlines the dataset details, competition framework, evaluation metrics, and final results, providing insights into the current state of rip current segmentation. We conclude with a discussion of key challenges, lessons learned from the submissions, and future directions for expanding RipSeg.
Background Artificial Intelligence (AI) is significantly transforming dermatology, particularly in early skin cancer detection and diagnosis. This technological advancement addresses a crucial public health issue by enhancing diagnostic accuracy, efficiency, and accessibility. AI integration in medical imaging and diagnostic procedures offers promising solutions to the limitations of traditional methods, which often rely on subjective clinical evaluations and histopathological analyses. This study systematically reviews current AI applications in skin cancer classification, providing a comprehensive overview of their advantages, challenges, methodologies, and functionalities. Methodology In this study, we conducted a comprehensive analysis of artificial intelligence (AI) applications in the classification of skin cancer. We evaluated publications from three prominent journal databases: Scopus, IEEE, and MDPI. We conducted a thorough selection process using the PRISMA guidelines, collecting 1,156 scientific articles. Our methodology included evaluating the titles and abstracts and thoroughly examining the full text to determine their relevance and quality. Consequently, we included a total of 95 publications in the final study. We analyzed and categorized the articles based on four key dimensions: advantages, difficulties, methodologies, and functionalities. Results AI-based models exhibit remarkable performance in skin cancer detection by leveraging advanced deep learning algorithms, image processing techniques, and feature extraction methods. The advantages of AI integration include significantly improved diagnostic accuracy, faster turnaround times, and increased accessibility to dermatological expertise, particularly benefiting underserved areas. However, several challenges remain, such as concerns over data privacy, complexities in integrating AI systems into existing workflows, and the need for large, high-quality datasets. AI-based methods for skin cancer detection, including CNNs, SVMs, and ensemble learning techniques, aim to improve lesion classification accuracy and increase early detection. AI systems enhance healthcare by enabling remote consultations, continuous patient monitoring, and supporting clinical decision-making, leading to more efficient care and better patient outcomes. Conclusions This comprehensive review highlights the transformative potential of AI in dermatology, particularly in skin cancer detection and diagnosis. While AI technologies have significantly improved diagnostic accuracy, efficiency, and accessibility, several challenges remain. Future research should focus on ensuring data privacy, developing robust AI systems that can generalize across diverse populations, and creating large, high-quality datasets. Integrating AI tools into clinical workflows is critical to maximizing their utility and effectiveness. Continuous innovation and interdisciplinary collaboration will be essential for fully realizing the benefits of AI in skin cancer detection and diagnosis.
Rund Mahafdah, Seifeddine Bouallegue, Ridha Bouallegue
The integration of Artificial Intelligence (AI) into e-learning has transformed conventional educational approaches, improving the learning process and maximizing student achievement. This study offers a thorough examination of how AI can be utilized to enhance e-learning results by employing advanced predictive methods and performance optimization strategies. The main goals consist of creating an AI-based framework to monitor and analyze student interactions, evaluating the influence of online learning platforms on student understanding using advanced algorithms, and determining the most efficient methods for blended learning systems. AI algorithms, known for their cognitive ability and capacity to learn, adapt, and make decisions, are employed to analyze and forecast student performance, thereby improving educational quality and outcomes. The practical results obtained by implementing machine learning and deep learning models, such as convolutional neural networks (CNN) and recurrent neural networks (RNN), show substantial enhancements in forecasting different performance metrics. This research highlights the ability of AI to develop adaptable, effective, and successful e-learning environments, promoting enhanced academic achievement and customized learning experiences. The findings demonstrate that CNN outperformed other deep learning and machine learning algorithms in terms of accuracy during the prediction phase, showcasing the advanced capabilities of AI in educational contexts. Portions of this text were previously published as part of a preprint (https://doi.org/10.21203/rs.3.rs-4724603/v1).
Background As Industry 4.0 debuted roughly a decade ago, it is now necessary to examine how it affects various aspects of the discipline. It is the responsibility of the education sector to guarantee that the next generation is equipped mentally, physically, and cognitively to face unforeseen challenges. Numerous educational institutions are outfitted with Industry 4.0 technology-based learning. Industry 4.0 fosters advancements in learning methodologies, especially for language enhancements. Learners may gain knowledge at their base, providing them an opportunity for independent study. The majority of subjects have been acquired through Industry 4.0. This research chapter explores the intersection of Industry 4.0 and education, specifically focusing on the SmallTalk AI tool. It investigates how technological and digital innovations within the context of Industry 4.0 can serve as powerful tools to enhance language learning outcomes. Methods This article presents a comprehensive analysis of statistical data and empirical evidence to support the positive impact of Industry 4.0 technology of SmallTalk on language acquisition particularly speaking. The study also determines the relationship among participants’ usage through the technology acceptance model (TAM). Furthermore, it examines the challenges and opportunities associated with integrating these innovations into language learning pedagogies, offering insights for educators and policymakers to harness the potential of Industry 4.0 in fostering language proficiency. The research employs quantitative analysis. The data obtained from educational institutions has been analyzed using the SPSS and AMOS software. Results The results indicate that Industry 4.0 has had an important effect on English language acquisition. This self-supported adaptable system of education facilitates effective student learning. This study also suggests that future research into the utility of Industry 4.0 be conducted elsewhere internationally.
The aim of this primer is to introduce the subject of knowledge engineering in a concise but synthetic way to develop the reader's intuition about the area.
The Communications Web site, http://cacm.acm.org, features more than a dozen bloggers in the BLOG@CACM community. In each issue of Communications , we'll publish selected posts or excerpts. twitter Follow us on Twitter at http://twitter.com/blogCACM http://cacm.acm.org/blogs/blog-cacm Mark Guzdial on a book that changed his thinking about teaching computer science, and Jiajie Zhang on the AI Revolution.
This is a commentary on the CP 2003 paper "Efficient cnf encoding of boolean cardinality constraints". After recalling its context, we outline a classification of Constraints with respect to their deductive power regarding General Arc Consistency (GAC).
This paper introduces the "grasp-ability test" as a "goodness" criteria by which to compare which explanation is more or less meaningful than others for users to understand the automated algorithmic data processing.