Cross-domain recommendation aims to improve recommendation accuracy in a target domain by leveraging auxiliary knowledge from a related source domain. In this study, we conduct a comprehensive empirical comparison between two prominent paradigms: Transfer Learning (TL) and Graph-Based (GB) methods. We construct an aligned book-movie dataset by combining the Book-Crossing and MovieLens-1M datasets, ensuring consistent preprocessing, hyperparameter tuning, and evaluation protocols. The TL methods, including DDTCDR and parameter-transferred NeuMF, focus on embedding reuse and fine-tuning across domains. In contrast, the GB methods, LightGCN and KGAT, integrate cross-domain edges and model higher-order semantic relations via graph propagation. Experimental results across multiple scenarios, including bidirectional transfer, cold-start items, and varying user-overlap ratios, show that GB approaches consistently outperform TL baselines, with KGAT achieving the best overall performance. In particular, GB methods demonstrate stronger robustness in low-overlap and cold-start settings, while TL approaches remain competitive and more efficient when overlap is high. We further analyze the trade-offs between interpretability, efficiency, and robustness, providing practical guidelines for model selection.
In this work, we propose and analyze a new fractional-order two-stage species model with recruitment, which combines a well-known integer-order two-stage species model with the Caputo fractional derivative, to discover memory effects on population dynamics. Firstly, the positivity and boundedness of solutions are investigated by using some standard comparison results. Next, a simple approach is utilized to study stability properties of the fractional-order model. This approach is based on the Lyapunov stability theory in combination with some nonstandard techniques for fractional dynamical systems. More clearly, we use general quadratic Lyapunov candidate functions and combine them with characteristics of quadratic forms associated with real matrices to establish the stability properties. Consequently, global asymptotic stability, uniform and Mittag-Leffler stability and, therefore, population dynamics of the proposed fractional-order model are analyzed rigorously. In addition, we extend Mickens' methodology to construct a dynamically consistent nonstandard finite difference (NSFD) scheme for the purpose of numerical simulation. It is proved that the NSFD scheme preserves the positivity and boundedness of the fractional-order model regardless of the values of the step size; moreover, it is also simple and efficient. Lastly, the theoretical results and advantages of the NSFD scheme are supported by illustrative numerical experiments.
PurposeThis study aims to examine the relationship between workplace incivility and volunteer employees' engagement (VEE), with a focus on the mediating role of emotional exhaustion. Additionally, the moderating role of resilient coping in this relationship is analysed.Design/methodology/approachThe data were collected through a survey using a convenience sampling method, in three waves, with a two-week interval between each wave, from a total of 258 employees working in Vietnam. To test the hypotheses, a moderated mediation model was estimated using the PROCESS macro with bootstrapping.FindingsThis study found that emotional exhaustion mediates the negative relationship between workplace incivility and VEE. Contrary to the theoretical expectations, resilient coping strengthens this mediated relationship.Practical implicationsThe findings highlight the need for particularly small and medium-sized enterprises (SMEs) to actively reduce workplace incivility and provide clearer volunteer role structures to prevent emotional exhaustion. Rather than relying solely on individual resilience, SMEs should strengthen structural support systems and protective mechanisms to sustain employee engagement.Originality/valueThis study extends prior research by examining emotional exhaustion as the underlying psychological mechanism and resilient coping as the boundary condition in the negative relationship between workplace incivility and VEE in SMEs. By integrating the conservation of resources theory and the transactional model of stress and coping, this study enhances theoretical understanding of how personal resources operate under interpersonal stress.
Growing interest in online education has led to considerable research on factors influencing students' learning outcomes, yet research on the relationship between teaching presence and student satisfaction and academic achievements in synchronous online English classes in developing countries is scanty. The current study aimed to address this gap by employing a quantitative design and Partial Least Squares Structural Equation Modeling (PLS-SEM) for data analysis. A total of 182 surveyed respondents from a private Vietnamese university, enrolling in various general English levels in synchronous online modes for at least one semester, completed a closed-ended five-point Likert scale questionnaire survey over a two-month period. The findings indicate that teaching presence was significantly associated with both extrinsic and intrinsic motivation, as well as student satisfaction in synchronous online English classes, although it does not directly influence academic performance. Extrinsic motivation showed a significant positive relationship with academic achievement, whereas intrinsic motivation shows no significant effect. Student satisfaction functions as a mediator between teaching presence and academic performance, with higher satisfaction correlating with improved academic outcomes. The research underscores the importance of teaching strategies that promote satisfaction and extrinsic motivation to bolster student success in online learning environments, specifically in the Vietnamese context, where online learning mode is burgeoning. Moreover, this study contributes to current research by providing additional empirical evidence on theoretical insights and applied applications for online English learning environments by focusing on the unique challenges and opportunities that synchronous online classes present in the Vietnamese educational system, thereby addressing a gap in research from developing countries.
This study investigates the impact of geopolitical risk (GPR) on stock market volatility and returns across six Southeast Asian countries. The motivation stems from the increasing frequency and severity of geopolitical events in recent years, particularly as emerging economies remain highly vulnerable to external shocks. Employing a panel vector autoregression model with exogenous variables (P-VARX), the analysis reveals that market volatility exhibits strong persistence, consistent with the well-documented phenomenon of volatility clustering in the financial literature. Contrary to much of the existing evidence from developed markets, the findings suggest that past geopolitical shocks tend to dampen current market volatility, reflecting a swift adjustment mechanism among investors in Southeast Asia. Furthermore, GPR exerts a positive influence on future stock returns. Impulse response functions and forecast error variance decompositions further elucidate the dynamic and asymmetric relationship between volatility and returns over time. These results offer novel empirical insights, indicating that while Southeast Asian markets are susceptible to geopolitical disturbances, they also demonstrate a remarkable capacity to absorb shocks and recover promptly. The study provides valuable implications for investors and policymakers in formulating effective risk management and investment strategies amidst rising global uncertainties.
