Universiti Tun Abdul Razak

This qualitative study explores the role of FinTech accelerators and incubators in fostering entrepreneurial innovation across seven major European cities: London, Berlin, Stockholm, Paris, Amsterdam, Zurich, and Barcelona. Through in-depth interviews with 16 entrepreneurs, mentors, and program managers, the research identifies the key support mechanisms, challenges, and outcomes associated with accelerator and incubator participation. The study reveals that financial support, mentorship, and networking are critical enablers for the growth of FinTech startups. However, regulatory complexities, talent acquisition, and market competition present persistent challenges. The findings highlight the significant impact of accelerators and incubators on startup scalability and market readiness. The research further emphasizes the need for harmonized regulatory frameworks and better talent development initiatives to support the growth of FinTech ecosystems in Europe. This study provides valuable insights for policymakers, accelerators, and entrepreneurs in enhancing the FinTech innovation ecosystem.

Olea europaea L. is an economically and ecologically significant species, for which accurate biomass estimation provides critical insights for artificial propagation, yield forecasting, and carbon sequestration assessments. Currently, research on biomass estimation for Olea europaea L. remains scarce, and there is a lack of efficient, accurate, and scalable technical solutions. To address this gap, this study achieved, for the first time, non-destructive estimation of Olea europaea L. biomass across individual tree to plot scales by integrating UAV-RGB (Unmanned Aerial Vehicle-Red-Green-Blue) imagery with the U2-Net model. This study initially developed allometric models for W-D-H, CA-D, and CA-H in Olea europaea L. (where W = biomass, D = ground diameter, H = tree height, and CA = canopy area). A single-parameter CA-based whole-plant biomass model was subsequently developed utilizing the optimal models. An innovative whole-plant biomass estimation model (UAV-RGB, U2-Net Total Biomass, UUTB) that combines UAV-RGB imagery with U2-Net at the sample-plot level was developed and assessed. The results revealed the following: (1) The model for Olea europaea L. aboveground biomass (AGB) was WA = 0.0025D1.943H0.690 (R2 = 0.912), the model for belowground biomass (BGB) was WB = 0.012D1.231H0.525 (R2 = 0.693), the model for CA-D was D = 4.31427C0.513 (R2 = 0.751), CA-H model was H = 226.51939C0.268 (R2 = 0.500). (2) The optimal AGB model for CA single-parameter was WA = 1.80901C1.181 (R2 = 0.845), and the model for BGB was WB = 1.25043C0.772 (R2 = 0.741). (3) The R2 of Olea europaea L. biomass, as estimated by CA derived from the U2-Net and UUTB models, was 0.855. This study presents the first integration of UAV-RGB imagery and the U2-Net model for biomass estimation in Olea europaea L., which not only addresses the research gap in species-specific allometric modeling but also overcomes the limitations of traditional manual measurement methods. The proposed approach provides a reliable technical foundation for accurate assessment of both economic yield and ecological carbon sequestration capacity.

Industry 4.0 is crucial in the industrial revolution, specifically focusing on interconnectivity, automation, machine learning, and real-time data. In this context, the Internet of Things (IoT) is critical, especially in smart environments, where machines are equipped with sensors and interconnected through web-enabled devices. Hence, individuals in this industry must constantly update their knowledge of IoT programming. Online learning is a flexible approach that allows learners to learn at their own pace and access courses from anywhere. However, existing educational websites for learning IoT programming are hindered by the lack of effective application of instructional design principles. Despite the importance of design thinking, there remains a paucity of evidence that educational websites incorporate critical components of effective learning paradigms. Additionally, these websites do not effectively integrate usability principles to make them easy to use and more pleasant. This study aimed to develop an educational website for learning IoT programming by integrating design thinking principles and the ADDIE model. The study uses a qualitative and quantitative approach in three phases to provide more significant insights. The iterative development process for design principles involves empathise, problem definition, ideating, prototyping, and testing. Meanwhile, the ADDIE model stages include analysis, design, development, implementation, and evaluation. Data were collected from ten learners through a semi-structured interview at the initial stage to identify problems with existing educational websites. Themes and sub-themes were developed from the interview responses using a thematic analysis. The findings were entered into an empathy map, and proposed solutions were used to develop the prototype. In the evaluation/testing stage, the prototype was evaluated by twenty-nine respondents to identify areas of improvement. The study findings were used to make recommendations to improve the prototype to enhance user experience, engagement, and effectiveness in learning IoT programming. The evidence from this study suggests that educational websites should optimize load times, improve mobile accessibility, enhance feedback mechanisms, provide real-time user support, and implement customization features.

The exponential growth of online financial transactions has significantly increased the vulnerability of banking and e-commerce systems to fraudulent activities, demanding intelligent, adaptive, and real-time detection mechanisms. This study presents an intelligent data-driven framework integrating machine learning, stream processing, and big data analytics for high-velocity transaction monitoring. The proposed architecture harnesses distributed data ingestion pipelines and stream-oriented processing engines to capture and analyze massive, continuously generated financial data streams with minimal latency. Feature engineering modules are designed to extract transactional, behavioral, and temporal features from heterogeneous data sources, while big data technologies such as Apache Spark and Kafka enable scalable real-time data handling. At the analytical core, the framework employs a hybrid ensemble of supervised and unsupervised learning models Random Forest (RF), Gradient Boosting (GBM), and Autoencoders to achieve robust detection of both known and novel fraud patterns. The models are trained on large-scale transactional datasets using feature selection and hyperparameter optimization strategies to ensure accuracy, interpretability, and generalization across dynamic environments. Streaming analytics and online learning components allow continuous model adaptation to evolving fraudulent behaviors without retraining from scratch. Experimental evaluations conducted on benchmark and synthetic datasets demonstrate the superior performance of the proposed framework in terms of detection rate, false-positive reduction, and computational efficiency compared with conventional batch-learning systems. The system achieves real-time throughput exceeding 50,000 transactions per second with sub-second decision latency, illustrating its suitability for deployment in large-scale financial ecosystems. In addition, explainable AI (XAI) modules are integrated to interpret model predictions and provide transparency in decision-making, thereby facilitating regulatory compliance and user trust. This research contributes to the ongoing advancement of intelligent financial security systems by merging data-driven learning with scalable stream analytics. The proposed framework offers a practical and generalizable solution for banks, payment gateways, and fintech platforms to identify fraudulent transactions proactively and adaptively in dynamic, high-velocity data environments. Future work will focus on integrating blockchain-based audit trails and federated learning for enhanced privacy and cross-institutional fraud intelligence sharing.

Background Hemolytic Disease of the Fetus and Newborn (HDFN) is the leading cause of anemia in fetuses and neonates, primarily due to maternal red blood cell alloantibodies that cross the placenta and attack fetal RBC with paternal antigens. The prevalence of these alloantibodies among pregnant women varies by country. This study aims to assess the local prevalence of clinically significant alloantibodies in pregnant women. Methods This retrospective study was conducted on total 6135 pretransfusion testing of pregnant women obtained from the hospital Laboratory Information System between 2020 and 2021 at the Blood Bank Unit of Hospital Canselor Tuanku Muhriz, University Kebangsaan Malaysia. Patients’ age, parity, transfusion history, sensitizing events and antibody screening results were collected. A descriptive analysis was conducted on the prevalence, frequency, specificity of clinically significant antibodies, and associated risk factors. Association of age and clinically significant alloantibodies were assessed using the chi-square test (p < 0.05). Results The prevalence of clinically significant alloantibodies was 0.8%. Anti-Mia was the most common alloantibody, followed by anti-E and anti-M. No significant association was detected between maternal age and clinically significant alloantibodies. Six primigravida patients had clinically significant alloantibodies—anti-M, anti-Mia, and anti-E—without any sensitizing events, suggesting they may be naturally occurring. All pregnant women should ideally receive red cell antibody screening during routine antenatal check-ups to address the risk of HDFN. However, in resource-limited settings with low prevalence of significant alloantibodies and HDFN, this may not be cost-effective. Targeted screening for high-risk populations could be a better alternative. BJMS, Vol. 24 No. 04 October’25 Page : 1138-1145

Enhancing the quality of education to achieve excellence and competitiveness is a top priority in the development of the national education system. This study employs a literature-based research method to explore key strategies for improving educational quality. The findings indicate that essential strategies include developing a competency-based curriculum, enhancing the quality of teaching staff, strengthening educational infrastructure, adopting modern learning technologies, and fostering collaboration between educational institutions and industry. Furthermore, the establishment of an academic culture that promotes innovation and creativity, along with the implementation of a rigorous quality assurance system, are critical factors in achieving sustainable improvement. In conclusion, improving the quality of education requires a holistic and sustainable approach involving all stakeholders in the educational ecosystem. This study contributes to the field of educational development by synthesizing key strategic elements that can serve as a practical framework for policymakers, educators, and institutional leaders in designing and implementing initiatives to enhance educational quality and competitiveness.

The contamination of endocrine disrupting compounds (EDCs) stands as an emerging environmental concern, leading to subsequent environmental and human exposure. A comprehensive analysis identified a total of 18 EDCs, including pharmaceuticals, hormones, and plasticizers, in Malaysian rivers classified into distinct pollution classes: "Clean", "Moderately Polluted", and "Polluted". The highest concentration of EDCs was observed in the "Moderately Polluted" Kim Kim River, notably containing 16.25µg/L of caffeine. Caffeine exhibited ubiquitous presence across all pollution classes, with the Klang River ("Polluted") showing the peak concentration at 13.62µg/L. In contrast, the "Clean" Kuantan River displayed the highest EDC concentration at 1.28µg/L of bisphenol A. All individual EDCs posed negligible ecological risks, with RQ values below 0.01 (RQm < 8.70 × 10⁻3). Similarly, negligible risks were observed for most EDCs under the worst-case scenario (RQex < 2.36 × 10⁻3). Noteworthy findings included the detection of previously undetected pharmaceuticals such as diphenhydramine on a global scale. Variability in the distribution of EDCs among river pollution classes exhibited statistically significant differences in their concentrations. The socioeconomic impact was evident, with gross domestic product (GDP) and population size positively influencing EDC concentrations, emphasizing the interconnected dynamics of urbanization, healthcare development, and pharmaceutical consumption. Additionally, the study identified negligible to low ecological risks associated with both individual and combined exposures to EDCs under general and worst-case scenarios. However, higher EDC risks were observed even in rivers classified as "Clean" or "Moderately Polluted", highlighting the need for more comprehensive monitoring strategies that account for emerging contaminants.

This article uses two different theoretical perspectives which are the welfare systems in the developing world, and institutional change dynamics to better understand the transition of Malaysia’s welfare system. Based on these theories in the Malaysian situation, it suggests that the welfare system of the country has been proved to have matured and changed. Over the course of economic growth, globalization and socio-productivist policies, this system has significantly transformed three times. Recent legal changes have added layers of complexity in which the government has embedded formal social programs onto long-standing informal welfare networks. This layering has produced complex results, with tensions between informal and formal welfare institutions that can be mutually undermining. The Malaysian case study contributes to our understanding of the characteristics and transformation of welfare regimes in developing countries in the sense that it fills a gap in the literature.

Artificial intelligence (AI) has demonstrated remarkable potential in transforming medical diagnostics across various healthcare domains. This paper explores AI applications in cancer detection, dental medicine, brain tumor database management, and personalized treatment planning. AI technologies such as machine learning and deep learning have enhanced diagnostic accuracy, improved data management, and facilitated personalized treatment strategies. In cancer detection, AI-driven imaging analysis aids in early diagnosis and precise treatment decisions. In dental healthcare, AI applications improve oral disease detection, treatment planning, and workflow efficiency. AI-powered brain tumor databases streamline medical data management, enhancing diagnostic precision and research outcomes. Personalized treatment planning benefits from AI algorithms that analyze genetic, clinical, and lifestyle data to recommend tailored interventions. Despite these advancements, AI integration faces challenges related to data privacy, algorithm bias, and regulatory concerns. Addressing these issues requires improved data governance, ethical frameworks, and interdisciplinary collaboration among healthcare professionals, researchers, and policymakers. Through comprehensive validation, educational initiatives, and standardized protocols, AI adoption in healthcare can enhance patient outcomes and optimize clinical decision-making, advancing the future of precision medicine and personalized care.