44197 research outputs found
Sort by
Benchmarking machine learning performance on small datasets
No full textArtificiële Intelligentie (AI) is onmisbaar in de samenleving. Ook KMO’s willen AI inzetten voor meer datagedreven en efficiënte operaties, maar hebben vaak te weinig data om modellen te trainen. Deze masterproef benchmarkt daarom drie state-of-the-art ML-algoritmen in een small-data scenario. De modellen worden getraind op datasets van 100–1000 voorbeelden, waarbij ook zes data-augmentatietechnieken en één dimensiereductiemethode (PCA) worden geëvalueerd. Experimenteel onderzoek met de CDC Diabetes- en Student Dropout-datasets omvat per trainingsgrootte tien steekproeven, waarna optimalisatie en evaluatie plaatsvinden met stratified KFold cross-validation. Performantiemetrieken zijn accuracy, precision, recall en F1-score. Populatiemodellen worden getraind op 70% van de volledige data om de small-data-resultaten te vergelijken met grote datasets. De resultaten tonen dat small-data-modellen vergelijkbare resultaten behalen als grote-data-modellen. Op de diabetesdataset leveren augmentatietechnieken nauwelijks verbetering in F1-score, terwijl PCA de gemiddelde F1-score met 20 procentpunten verhoogt. Op de studentendataset verbeteren SVMSMOTE, BorderlineSMOTE, SMOTE en ADASYN de F1-score van RF- en XGB-modellen licht (ongeveer 1%) ten opzichte van de baseline; bij LogReg is geen enkele augmentatietechniek effectief
The inclusive potential of activity-based working: The case of disability
No full textOrganisations are increasingly adopting activity-based working, replacing assigned desks and private offices with open, shared workspaces while shifting some tasks to employees' homes. Management promotes these changes through explicit 'efficiency-gain claims' and subtler promises of flexibility and de-hierarchisation, thus making an implicit 'inclusion-gain claim'. Drawing on 35 interviews, this study challenges the inclusion-gain assumption by examining activity-based working through the lens of visibility. The case of disabled workers reveals how the nature of disability and impairments complicate visibility in diverse and often ambiguous ways. Our findings show that while perceptions of diversity have increased with activity-based working, meaningful inclusion relies on universal flexibility and higher accessibility standards. Yet, disabled workers' exclusion from the design phase resulted in retrofitting and exceptional territoriality, threatening inclusion. We contribute to the flexibility-visibility debate by showing that flexible organisational spaces not only influence the visibility of employees in terms of work recognition through spatial dispersion but also shape the visibility of embodied differences through spatial othering. Finally, this study reconceives disabled individuals and their allies not as passive users but as active re-designers of ableist workspaces, redefining visibility as a socially constructed, contested process shaped by the spatial and organisational structures of work.Funding
The authors disclosed receipt of the following financial support for the research, authorship and/orpublication of this article: This research was funded by the Fonds de la Recherche Scientifique(F.R.S. – FNRS). The first author is a Research Fellow of the F.R.S. – FNRS
Impact of future climate change and land-use change on habitat suitability for a long-distance avian migrant under diverse socioeconomic-emission scenarios
No full textClimate change and the alteration of land-use and land-cover (LULC) on a global scale presents a notable threat to biodiversity, with a particular impact on long-distance migratory birds. Nonetheless, our comprehension of how these changes specifically affect these avian migrants remains limited. To address this knowledge gap, species distribution models offer a valuable tool as they allow for projecting habitat suitability and provide insights into a species' ecological responses to changing environmental conditions. In this study, we utilize modelled climate and LULC data spanning from 2015 to 2100, coupled with GPS tracking data and open-source occurrence data of European Nightjars (Caprimulgus europaeus), to assess how suitable habitat might be affected by changes in climate and LULC under different emission and socio-economic scenarios. Our models show that climate change alone may expand suitable habitat, particularly under high-emission scenarios, while LULC change generally exerts a constraining effect across annual cycle stages. While some scenarios indicate overall stability in the amount of suitable habitat, our spatially explicit results reveal that these entail substantial redistribution of suitable habitat which may disrupt migration patterns and increase energetic costs. Our findings underscore the importance of explicitly considering spatial and temporal aspects in assessments of global change impacts on migratory species and highlight the need for targeted conservation interventions in critical stopover zones to safeguard the species' future amidst global environmental change.Funding
This work was supported by a BOF-mandate at Hasselt University and University of Namur (R-9983), the FWO - research foundation Flanders (12T3922N), Forestry England, and the British Birds Charitable Trust.
Acknowledgements
The authors wish to thank Kathe ¨ Robert, Gunter Flipkens and Marc Evens for help during fieldwork, Hans Van Dyck, Luc Lens, Marcel Eens, Barbara Helm, M. Constantine, Koen Thijs, Annie Loenders, Karen Vanmarcke, and Fien and Fleur Evens for support. Belgian permissions were granted by the Belgian military, Agency for Nature and Forest and Royal Belgian Institute for Natural Sciences. This research used resources of the "Plateforme Technologique de Calcul Intensif (PTCI)" (http://www.ptci.unamur.be) located at the University of Namur, Belgium, which is supported by the FNRS-FRFC, the Walloon Region, and the University of Namur (Conventions No. 2.5020.11, GEQ U.G006.15, 1610468, RW/GEQ2016 et U.G011.22). The PTCI is member of the "Consortium des Equipements ´ de Calcul Intensif (CECI)" ´ (http://www.ceci-hpc.be). Further resources and services used in this work were provided by the VSC (Flemish Supercomputer Center), funded by the Research Foundation - Flanders (FWO) and the Flemish Government
Geautomatiseerde en gedigitaliseerde inschatting van competenties
No full textDeze masterproef beschrijft de ontwikkeling van een virtual reality (VR)-gebaseerd systeem om de objectiviteit en schaalbaarheid van competentiebeoordelingen voor medewerkers van Bewel, een organisatie in Limburg die personen met arbeidsproblemen ondersteunt, te verbeteren. De huidige werkwijze steunt op handmatige, observatiegestuurde tests die subjectief worden beoordeeld door begeleiders, wat leidt tot variabele uitkomsten en een hoge tijdsinvestering. Als oplossing van deze beperkingen is in Unity, gebruikmakend van de OpenXR-standaard, een VR-omgeving ontwikkeld om drie screeningopdrachten te simuleren: inhoudsinspectie van een doos, het gebruiken van een zeefdrukmachine en het stapelen van emmers volgens een gegeven patroon. Het systeem registreert gebruikersacties en simuleert interacties via VR-handtracking. Daarnaast werd een begeleidersdashboard ontwikkeld waarmee testresultaten per opdracht gevisualiseerd en geëvalueerd kunnen worden. Dit dashboard ondersteunt objectieve metingen door de gemeten data te visualiseren en maakt het mogelijk om scores automatisch te genereren. Uit de evaluatie met maatwerkers en begeleiders blijkt dat de VR-toepassing zowel technisch als functioneel haalbaar is. De resultaten tonen aan dat het systeem de subjectiviteit in de beoordeling vermindert, doordat het de zekerheid van de beoordelaars verhoogt wat resulteert in een meer objectieve en uniforme evaluatie. De case study bevestigt het potentieel van VR als aanvullend screeningsinstrument binnen maatwerkbedrijven
Unveiling a Two-Electron Reaction Pathway for Electrocatalytic CO2 Reduction on Boron-Doped Diamonds: A First-Principles Calculation
No full textConverting excessive CO2 molecules into formic acid (HCOOH) as a liquid fuel and hydrogen storage carrier using a sustainable electrochemical method has received enormous attentions. However, the reaction mechanism during this two-electron reaction pathway is still controversial. Inspired by the high selectivity toward HCOOH on the boron-doped diamond (BDD) electrode, this work calculates the adsorption of the CO2 molecule and first two-electron reaction pathway on BDD with different B doping configurations by the density functional theory method. The results show that CO2 molecules are more readily adsorbed on the surface B doping sites with charge transfer between B-O bonding. And the total overpotential of the first two-electron reaction pathway displays a Volcano relationship with the Gibbs energy of the *CO2-*COOH step. The partially sp2-C hybridized (111) (2 x 1) configuration exhibits the lowest overpotential of 0.81 eV and the best CO2 reduction performance toward the HCOOH product. Furthermore, the dynamic kinetics of the *CO2-*COOH step is investigated by the climbing image-nudged elastic band method under the external electric field. The negative electric field of -0.4 eV/& Aring; promotes the adsorption of CO2 and *H but inhibits the migration of *H with an energy barrier of 4.34 eV. This work elucidates the decision factor of high selectivity toward the HCOOH product on the BDD electrode and provides a comprehensive understanding of two-electron reaction pathway mechanisms.This work was funded by the National Natural Science Foundation of China (Grant No. 52402063)
At the Frontline of Healthcare: Understanding the Impact of Frontline Service Technologies
No full textFrontline service technology (FST), defined as any combination of hardware, software, information systems, and networks that facilitates value co-creation between a healthcare provider and healthcare customer by optimizing processes and outcomes across the healthcare frontline, is rapidly transforming the frontline of healthcare. These technologies hold significant potential to address the mounting challenges faced by the healthcare industry, yet they also reshape the roles and experiences of both healthcare customers and providers. Although research on FSTs is expanding, much of it lacks service-related insights. This dissertation broadens the lens through which FSTs are evaluated by investigating their effectiveness in reaching healthcare objectives and their impact on healthcare customers and providers, through a service-oriented lens. Each of the six chapters contributes to these objectives through distinct methodological approaches.
Chapter 1 outlines the transformation of the organizational frontline and introduces FSTs in the context of healthcare services. It also details the structure of the dissertation and its intended contributions. Chapter 2 investigates the effectiveness of FSTs in advancing the Quintuple Aim of healthcare: improving health outcomes, reducing costs of care, enhancing healthcare customer and provider experiences, and promoting health equity. A systematic literature review reveals that while FSTs improve health outcomes, reduce costs of care, and enhance healthcare customer experience, their impacts on provider experience and health equity are neglected. Existing research adopts a predominantly clinical lens, underutilizing service-related insights. The chapter concludes with a research agenda identifying five key areas where service researchers can contribute.
Chapters 3 and 4 shift the focus to the impacts on FST users. Grounded in the Job Demands–Resources (JD-R) theory, both chapters examine how FSTs reshape users’ roles by influencing the demands they encounter and the resources they can draw on. Chapter 3 uses a qualitative study to explore the bright and dark sides of FSTs. Two remote monitoring FSTs are studied: Pregnancy Remote Monitoring and Elderly Remote Monitoring, considering perspectives from both customers and providers. Findings reveal distinct role demands and role resources during FST usage, and psychological capital (PsyCap) emerges as a critical personal resource. The chapter emphasizes shared patterns and differences across user groups, highlighting the need for tailored support strategies.
Chapter 4 extends these insights through scenario-based studies that empirically test the JD-R theory. Using partial least squares structural equation modeling, it investigates how role demands, role resources, and PsyCap influence users’ well-being, satisfaction, and word-of-mouth intentions. The results show that PsyCap significantly enhances motivation, satisfaction, and word-of-mouth; role demands increase strain but have no impact on motivation; and role resources are strongly linked to key outcomes in several user groups. Furthermore, well-being emerges as a key driver of satisfaction and word-of-mouth.
Chapter 5 introduces WATCH (Well-Being Assessment of Technology Concerning Healthcare), a research-driven application developed to help organizations monitor and optimize the impact of FSTs. Drawing on Importance-Performance Analysis, the tool provides actionable insights for improving user well-being and organizational outcomes.
Chapter 6 concludes the dissertation by reflecting on key lessons, acknowledging limitations, and offering directions for future research
Unlocking Tomorrow’s Classrooms: Attitudes and Motivation Toward Data-Based Decision-Making in Teacher Education
No full textIn today's increasingly data-driven educational landscape, teachers are expected to use data to inform instructional decisions. However, effective data use depends not only on statistical competence but also on motivation, attitudes, and academic self-concept. This study examines how these factors influence student teachers' readiness to engage with standardized assessment data. A survey of 164 Flemish primary education student teachers assessed their motivation, attitudes toward data use, and academic self-concept. Cluster analysis identified four distinct profiles, ranging from highly competent yet disengaged users to low-performing but externally motivated individuals, highlighting significant variability in data engagement. A pre-and post-test study design involving an e-course on basic statistical concepts demonstrated that targeted instruction can enhance perceived competence, particularly in areas such as box plot interpretation. Findings suggest that technical training alone is insufficient to promote sustained data use; fostering intrinsic motivation, positive attitudes, and a strong academic self-concept is essential for long-term engagement with data.This study is part of a research project on the use of feedback from standardized tests within the Steunpunt Centrale Toetsen in Onderwijs, funded by the Flemish Ministry of Education and Training (Flanders, Belgium)
The impact of millennial CEOs on family firm internationalization: an attention-based view
No full text
Tracking Through-Thickness Dynamics in Lithium-Sulfur Batteries via Operando Raman Imaging and a Purpose-Built Cell
No full textMost existing operando methods lack the spatial resolution or measurement access required for cross-sectional analysis of battery electrodes. Here, we present a custom-designed operando electrochemical cell, enabling real-time, through-thickness investigations. Applied to a lithium-sulfur battery, this setup reveals gradual and spatially non-uniform sulfur utilization, dissolution, and reprecipitation. Depth-resolved measurements indicate highly nonuniform sulfur utilization, with earlier sulfur depletion near the current collector. Accessing the electrode cross-section provides critical insights into reaction heterogeneity and dynamics that are inaccessible via conventional measurements. This offers a deeper understanding of battery behavior and facilitates the design of cells with improved performance and durability.This work was supported by SIM (Strategic Initiative Materials in Flanders) and VLAIO (Flemish Government Agency Flanders Innovation and Entrepreneurship) within the SBO project “FuGels” (Grant HBC.2021.0016) under the SIM research program “SIMBA—Sustainable and Innovative Materials for Batteries.