Abstract
The branch-and-bound (B&B) procedure is one of the most frequently used methods for solving the resource-constrained project scheduling problem (RCPSP) to obtain optimal solutions and has a rich history in the academic literature. Over the past decades, various variants of this procedure have been proposed, each using slightly different configurations to search for the optimal solution. While most of the configurations perform relatively well for many problem instances, there is, however, no known universal best B&B configuration that works well for all problem instances. In this work, we propose two problem transformation-based machine learning classification methods (binary relevance and classifier chains) to automatically detect the best-performing branch-and-bound configuration for the resource-constrained project scheduling problem. The proposed novel learning models aim to find the relationship between the project characteristics and the performance of a specific B&B configuration. With this obtained knowledge, the best-performing B&B configurations can be predicted, resulting in a better solution. A comprehensive computational experiment is conducted to demonstrate the effectiveness of the proposed classification models and the performance improvements over three categories of methods from the literature, including the latest branch-and-bound configurations, the state-of-the-art classification models in project scheduling, and commonly used clustering algorithms in machine learning. The results show that the proposed classification models can enhance solution quality for the RCPSP without changing the core components of existing algorithms. More specifically, the classifier chains method, when combined with the Back-Propagation Neural Network algorithm, achieves the best performance, outperforming binary relevance, which demonstrates the impact of label correlation on the performance. The experiments also demonstrate the merits of the proposed model in improving the robustness of the solutions. Furthermore, these findings not only highlight the potential of the classification models in detecting best-performing B&B configurations, but also emphasize the need for future work and development to further improve the performance and applicability of these models.

Abstract
Olive oil production is a noteworthy economic activity in multiple places worldwide. Due to environmental degradation and lack of resources with population growth, there is a global tendency for more sustainable and efficient practices, driving the implementation of more responsible agricultural and industrial systems. This paper aims to develop an intelligent system architecture focused on optimizing the production of olive oil, improving product quality, reducing operational waste, and maximizing the efficient use of natural resources. Through the use of Industrial Internet of Things (IIoT) technologies, the proposed solution aims to monitor and control the parameters of olive oil production automatically. In addition, the study addresses sensors already used in the market and existing systems to compare and seek improvements. The proposed architecture contains three layers: device, edge, and cloud computing layer, which are integrated and enable the implementation of a scalable and complete solution that allows real-time visualization and control of the production process.

Abstract
The increasing integration of distributed energy generation into the electrical grid has led to changes in the structure and organization of energy markets over the past years. Market trading has become increasingly demanding due to the different types of production profiles. A forecast of the total production of all assets is made to bid for energy. Whenever there are differences between the forecast and the actual produced energy, a deviation occurs, which is assigned to the agent responsible for its settlement. This article proposes the application of a linear regression algorithm supported by a clustering method to forecast energy production. Based on the historical production profile of the installations in each cluster, it is possible to predict the production pattern for a period with no available data, thus standardizing this data for other assets belonging to the same cluster.

Abstract

Abstract
Driving simulators are essential tools for training, education, research, and scientific experimentation. However, the diversity and quality of virtual environments in simulations is limited by the specialized human resources availability for authoring the content, leading to repetitive scenarios and low complexity of real-world scenes. This work introduces a pipeline that can process text-based narratives outlining driving experiments to procedurally generate dynamic traffic simulation scenarios. The solution uses Retrieval-Augmented Generation alongside local open-source Large Language Models to analyse unstructured textual information and produce a knowledge graph that encapsulates the world scene described in the experiment. Additionally, a context-based formal grammar is generated through inverse procedural modelling, reflecting the game mechanics related to the interactions among the world entities in the virtual environment supported by CARLA driving simulator. The proposed pipeline aims to simplify the generation of virtual environments for traffic simulation based on descriptions from scientific experiment, even for users without expertise in computer graphics. © 2025 by SCITEPRESS–Science and Technology Publications, Lda.

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.