Abstract
Data-based approach for diagnosis of thyroid disorders is still at its early stage. Most of the research outcomes deal with binary classification of the disorders, i.e. presence or not of some pathology (cancer, hyperthyroidism, hypothyroidism, etc.). In this paper we explore deep learning (DL) models to improve the multi-class diagnosis of thyroid disorders, namely hypothyroid, hyperthyroid and no pathology thyroid. The proposed DL models, including DNN, CNN, LSTM, and a hybrid CNN-LSTM architecture, are inspired by state-of-the-art work and demonstrate superior performance, largely due to careful feature selection and the application of SMOTE for class balancing prior to model training. Our experiments show that the CNN-LSTM model achieved the highest overall accuracy of 99%, with precision, recall, and F1-scores all exceeding 92% across the three classes. The use of SMOTE for class balancing improved most of the model’s performance. These results indicate that the proposed DL models not only effectively distinguish between different thyroid conditions but also hold promise for practical implementation in clinical settings, potentially supporting healthcare professionals in more accurate and efficient diagnosis. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

Abstract
The Vision Transformer (ViT) architecture has emerged as a potential game-changer in computer vision, offering scalability and global attention that have generated considerable interest in recent years. Its adaptability has fueled enthusiasm for its application. This work investigates the boundaries of the architecture, focusing on developing new techniques targeting explicitly complex tasks, such as medical imaging datasets, which often exhibit high variability, class imbalance, and limited sample sizes. We propose a set of mixed regularisation and augmentation techniques to enhance the performance of models. These include a novel loss function and a smoothly differentiable activation function, leading to more stable training and model performance. The results show that incorporating these techniques improves model performance and training convergence.

Abstract
Uncertainty is pervasive in modern manufacturing settings. In order to cope with unexpected events, scheduling decisions are commonly taken resorting to dispatching rules, which are reactive in nature. However, rule performance varies according to shop utilisation and due date allowance, which often change in dynamic real-world job shops. Therefore, this paper explores systems that select dispatching rules as conditions change over time, namely periodic and real-time dispatching rule selection systems, which are based on supervised learning and reinforcement learning algorithms, respectively. These types of systems have been proposed in the past but have been further improved in this work by carefully selecting the most relevant state features and dispatching rules. Moreover, by testing both approaches on the same instances, it was possible to compare them and determine the most advantageous one. After the tests, which included a wide array of job shop instances, both periodic and real-time systems outperformed state-of-the-art dispatching rules by over 10% tardiness-wise. Nonetheless, the periodic rule selection approach was more robust across all tests than the real-time approach. These results demonstrate that there is a real incentive for managers to adopt dispatching rule selection systems.

Abstract

Abstract
Knowledge sharing and team dynamics are essential elements of entrepreneurial success, especially in teams that operate in innovative environments. This chapter explores how participation in an incubation program influences the formation and development of entrepreneurial team identity. It aims to understand the dynamics involved in creating entrepreneurial teams, the practices of knowledge sharing, and the role digital technologies play in supporting and sustaining these processes. The study focuses on teams that completed the CEU iLab Incubation Program, with data gathered through in-depth semistructured interviews from twenty-five entrepreneurs across various startups. Five cases, involving entire entrepreneurial teams, were central to this research. The findings offer valuable insights for enhancing incubation programs, promoting entrepreneurial identity formation, and improving the success of new ventures. These insights are beneficial for both scholars and practitioners in the entrepreneurship field. © 2025 by IGI Global Scientific Publishing. All rights reserved.

Abstract
Integrating Industry 4.0 and Quality 4.0 optimises manufacturing through IoT and ML, improving processes and product quality. The primary challenge involves identifying patterns in computer numerical control (CNC) machining time-series data to boost manufacturing quality control. The proposed solution involves an experimental study comparing one-class and binary classification algorithms. This study aims to classify time-series data from CNC turning machines, offering insight into monitoring and adjusting tool wear to maintain product quality. The methodology entails extracting spectral features from time-series data to train both one-class and binary classification algorithms, assessing their effectiveness and computational efficiency. Although certain models consistently outperform others, determining the best performing is not possible, as a trade-off between classification and computational performance is observed, with gradient boosting standing out for effectively balancing both aspects. Thus, the choice between one-class and binary classification ultimately relies on dataset's features and task objectives.