Abstract
Accurate evaluation of forecasting models is essential for ensuring reliable predictions. Current practices for evaluating and comparing forecasting models focus on summarising performance into a single score, using metrics such as SMAPE. We hypothesize that averaging performance over all samples dilutes relevant information about the relative performance of models. Particularly, conditions in which this relative performance is different than the overall accuracy. We address this limitation by proposing a novel framework for evaluating univariate time series forecasting models from multiple perspectives, such as one-step ahead forecasting versus multi-step ahead forecasting. We show the advantages of this framework by comparing a state-of-the-art deep learning approach with classical forecasting techniques. While classical methods (e.g. ARIMA) are long-standing approaches to forecasting, deep neural networks (e.g. NHITS) have recently shown state-of-the-art forecasting performance in benchmark datasets. We conducted extensive experiments that show NHITS generally performs best, but its superiority varies with forecasting conditions. For instance, concerning the forecasting horizon, NHITS only outperforms classical approaches for multi-step ahead forecasting. Another relevant insight is that, when dealing with anomalies, NHITS is outperformed by methods such as Theta. These findings highlight the importance of evaluating forecasts from multiple dimensions.

Abstract
Background and aims: Reward-based eating reflects hedonic drivers of intake, including loss of control, diminished satiety, and preoccupation with food. We translated, adapted and studied the psychometric properties of the 13- and 5-item Reward-Based Eating Drive Scale (RED), for Portugal, Poland and Italy. Methods: A cross-cultural study was conducted with higher education students and general population samples (n = 1999). After translation and cultural adaptation, the RED was administered with food craving items, and collection of sociodemographic and anthropometric data. Factorial structure and measurement invariance were tested using confirmatory factor analysis (CFA), internal consistency with Cronbach's alpha, and convergent validity via correlations with BMI and cravings. Results: CFA supported the expected structures of the RED-13 (three factors) and RED-X5 (unifactorial), with configural and metric invariance across countries and groups. Only partial scalar invariance was achieved for both versions. The RED-13 showed good to excellent internal consistency for total scores (0.868 <= alpha <= 0.906), with acceptable to good reliability for Loss of control (0.769 <= alpha <= 0.821), lower values for Lack of satiety (0.655 <= alpha <= 0.723), and good to excellent consistency for Preoccupation with food (0.881 <= alpha <= 0.918). The RED-X5 showed acceptable internal consistency (0.737 <= alpha <= 0.811) and correlated strongly with RED-13 (r = 0.949, p < 0.001). Both correlated positively with BMI and food cravings. Age, sex, and country had small to medium multivariate effects on RED scores. Conclusions: The RED-13 and RED-X5 showed good psychometric properties in Portugal, Poland, and Italy, with the RED-13 providing a multifactorial assessment and the RED-X5 offering a brief alternative.

Abstract
Small, medium, and large organizations collect vast amounts of data with the expectation of using it to generate commercial value. Machine learning is a powerful tool for extracting valuable insights from this data and serves as a pivotal sales strategy for companies to maximize profits. This paper seeks to analyze sales data and discern patterns in sales among products that exhibit similarities, such as boxes and bags. In order to achieve this goal, was used unsupervised learning methods that allow the segmentation of groups, specifically Principal Component Analysis (PCA), k-means algorithms, and hierarchical clustering. PCA was used to identify correlated variables and find hidden patterns in the data, particularly pertaining to product families with similar sales. Elbow, Silhouette, and 30 indices methods were applied to determine the optimal number of clusters. Based on these results, it was determined the optimal number of clusters. Validation methods were employed to identify the clustering algorithm exhibiting the best performance. Stability measures evaluated the consistency of the clusters, while the cophenetic coefficient aided in determining the most effective data grouping method. After validation, the clustering algorithms were implemented. The results indicated that all clustering algorithms effectively segmented the data, with particular emphasis on the performance of the k-means algorithm. This study identified product groups with similar sales patterns and key products that impact the company’s global sales. Multivariate analysis provided a deeper understanding of sales dynamics, enabling the company to implement targeted marketing strategies and optimize resource allocation to boost bag and box sales in Portugal and other countries. © 2025 Elsevier B.V., All rights reserved.

Abstract
Our team previously introduced an innovative concept for an "invisible" Electrocardiography (ECG) system, incorporating electrodes and sensors into a toilet seat design to enable signal acquisition from the thighs. Building upon that work, we now present a novel dataset featuring real-world, single-lead ECG signals captured at the thighs, offering a valuable resource for advancing research on thigh-based ECG for cardiovascular disease assessment. To our knowledge, this is the first dataset of its kind. The tOLIet dataset comprises 149 ECG recordings collected from 86 individuals (50 females, 36 males) with an average age of 31.73 ± 13.11 years, a mean weight of 66.89 ± 10.70 kg, and an average height of 166.82 ± 6.07 cm. Participants were recruited through direct contact with the Principal Investigator at Centro Hospitalar Universitario de Lisboa Central (CHULC) and via clinical consultations conducted at the same institution. Each recording includes four differential signals acquired from electrode pairs embedded in the toilet seat, with reference signals obtained from a standard 12-lead hospital ECG system.

Abstract

Abstract
This paper contributes to developing a Method for Creating Persona Templates (MCPT), addressing a significant gap in user-centred design methodologies. Utilising qualitative data collection and analysis techniques, MCPT offers a systematic approach to developing robust and context-oriented persona templates. MCPT was created by applying the Design Science Research (DSR) methodology, and it incorporates multiple iterations for template refinement and validation among project stakeholders; all of the proposed steps of this method were based on theoretical contributions. Furthermore, MCPT was tested and refined within a real-life R&D project focusing on developing a digital platform e-marketplace for short agrifood supply chains in two iteration cycles. MCPT fills a critical void in persona research by providing detailed instructions for each step of template development. By involving the target audience, users, and project stakeholders, MCPT adds rigour to the persona creation process, enhancing the quality and relevance of personae casts. This paper contributes to the body of knowledge by offering an initial proposal of a comprehensive method for creating persona templates within diverse projects and contexts. Further research should explore MCPT's adaptability to different settings and projects, thus refining its effectiveness and extending its utility in user-centred design practices.