Search Results (758)

Accurate credit risk assessment for small firms and farmers is crucial for financial stability and inclusion; however, many models still rely on binary default labels, overlooking the continuum of borrower vulnerability. To address this, we propose Transformer–LightGBM–Stacked Logistic Regression (TL-StackLR), a hybrid stacking framework for multi-class loan default prediction. The framework combines three learners: a Feature Tokenizer Transformer (FT-Transformer) for feature interactions, LightGBM for non-linear pattern recognition, and a stacked LR meta-learner for calibrated probability fusion. We transform binary labels into three risk tiers, Low, Medium, and High, based on quantile-based stratification of default probabilities, aligning the model with real-world risk management. Evaluated on datasets from 3045 firms and 2044 farmers in China, TL-StackLR achieves state-of-the-art ROC-AUC scores of 0.986 (firms) and 0.972 (farmers), with superior calibration and discrimination across all risk classes, outperforming all standalone and partial-hybrid benchmarks. The framework provides SHapley Additive exPlanations (SHAP) interpretability, showing how key risk drivers, such as income, industry experience, and mortgage score for firms and loan purpose, Engel coefficient, and income for farmers, influence risk tiers. This transparency transforms TL-StackLR into a decision-support tool, enabling targeted interventions for inclusive lending, thus offering a practical foundation for equitable credit risk management. Full article

The lack of explicit negative labels issue is a prevalent challenge in numerous domains, including CV, NLP, and Recommender Systems (RSs). To address this challenge, many negative sample completion methods are proposed, such as optimizing sample distribution through pseudo-negative sampling and confidence screening in CV, constructing reliable negative examples by leveraging textual semantics in NLP, and supplementing negative samples via sparsity analysis of user interaction behaviors and preference inference in RS for handling implicit feedback. However, most existing methods fail to adequately address the Missing-Not-At-Random (MNAR) nature of the data and the potential presence of unmeasured confounders, which compromise model robustness in practice. In this paper, we first formulate the prediction task in RS with implicit feedback as a positive-unlabeled (PU) learning problem. We then propose a two-phase debiasing framework consisting of exposure status imputation, followed by debiasing through the proposed doubly robust estimator. Moreover, our theoretical analysis shows that existing propensity-based approaches are biased in the presence of unmeasured confounders. To overcome this, we incorporate a robust deconfounding method in the debiasing phase to effectively mitigate the impact of unmeasured confounders. We conduct extensive experiments on three widely used real-world datasets to demonstrate the effectiveness and potential of the proposed methods. Full article

Advancements in data storage and data processing technologies has compelled higher education institutions to optimise the use of their data. Many universities globally have begun to implement learning analytics at their institutions to better understand and improve teaching and learning. African higher education institutions have been slow to implement learning analytics despite the continued accumulation of digital data. The research related to this study presents a dataset of Information Systems and Technology (IS&T) students from the University of KwaZulu-Natal, a South African university. The dataset comprises approximately 14,000 registered student records from 10 IS&T courses, primarily consisting of demographic data, academic performance (including past IS&T courses and school records), and Learning Management System (LMS) interaction data. The dataset exhibits an imbalance, characterised by a higher proportion of students who have successfully completed courses compared to those who have not. The dataset will be of interest to researchers engaged in learning analytics application studies, including early pass/fail prediction and grade classification, as well as those who want to test their techniques on a real-world dataset. Full article

Purpose: To describe structural and systemic barriers to ophthalmic care experienced by underserved patients, particularly those facing language obstacles, immigration-related constraints, limited insurance coverage, financial hardship, and navigation challenges in an urban setting, and to examine these barriers through a complexity-informed lens. Methods: We conducted a narrative literature review focused on healthcare disparities, patient navigation, complexity in care delivery, and time-sensitive prioritization frameworks in ophthalmology. Findings were integrated with case vignettes drawn from Eyes on Wheels (EOW), a mobile eye care initiative that provides no-cost examinations at Federally Qualified Health Centers (FQHCs) and free clinics. Cases were identified through routine clinical documentation and used to illustrate how structural barriers described in the literature manifest in real-world care pathways. Results: Three recurring system-level issues were identified across EOW encounters: (A) misclassification of medically necessary, time-sensitive ophthalmic care as “non-urgent”; (B) patient disengagement driven by cumulative structural and logistical barriers; and (C) failures that arise when the healthcare system, functioning as a complex adaptive system (CAS), is unable to adapt to patients’ and systems’ changing circumstances. A review of the literature confirmed that these patterns reflect widely documented challenges faced by underserved urban populations. Three EOW case vignettes, selected from seven patients identified in 2024, are presented as illustrative examples of these systemic patterns. Conclusions: Addressing inequities in eye care requires an approach that recognizes how many parts of the healthcare system interact and affect a patient’s ability to receive timely treatment. Vision loss is often the preventable result of systems that are rigid, fragmented, or unable to adapt to a patient’s circumstances. Improving outcomes will require flexible care models, such as mobile clinics, paired with strong institutional support, patient-centered navigation, and consistent assessment of social needs and barriers to care. Sustained progress will depend on collaboration across organizations, adaptable leadership, and policies that respond to the real-world situations in which patients live. Full article

Many countries around the world require some or all schools to develop and implement a school improvement plan (SIP), which is a tool intended to guide the identification of school-specific needs for improvement along with a series of priorities, goals, and actions to help address those needs. Yet, the literature on this topic remains rather sparse. In this article, we conducted a systematic review of the international literature on SIPs published from 2010 through 2025, identifying 62 relevant articles for analysis. We organized this review’s findings around six areas related to SIP development and implementation: assessing current conditions, determining needs, setting direction, organizing resources, taking action, and evaluating progress. Findings suggest that while divergences exist between contexts with respect to these six areas, there are considerable convergences in how educators and others conceptualize and interact with SIPs. We close with recommendations for future research that both strengthens and broadens the extant literature on SIPs. Full article

The year 2025 marks two significant milestones for aflatoxin research: 65 years since aflatoxin was first identified in 1960, and 50 years of focused research on preharvest aflatoxin contamination since it was first recognized in 1975. Studies in the 1970s revealed that A. flavus could infect crops like maize and produce aflatoxin in the field before harvest and made it possible to investigate the potential genetic resistance in crops to mitigate the issues. Tremendous efforts have been made to learn about the process and regulation of aflatoxin production along with interactions between A. flavus and host plants as influenced by environmental factors. This has allowed for the breeding of more resistant crops and investigations into the underlying genetic and genomic components of resistance mechanisms in crops like maize and peanut. However, despite decades of studies, many questions remain. One established “dogma” is that drought stress, especially when combined with high temperatures, is the single greatest contributing factor to preharvest aflatoxin contamination and is a perennial risk faced throughout the major agricultural production regions of the world. Although there are many reviews summarizing the decades’ long wealth of information about A. flavus, aflatoxin biosynthesis, management and host plant resistance, there are few reports that put the spotlight on why aflatoxin contamination is exacerbated by drought stress, which places plants under severe physiological stress and weakens immune systems. Therefore, here we will focus on three major areas of research in maize: the “living embryo” theory and host resistance mechanisms, the “Key Largo hypothesis” and the causes of drought-exacerbated aflatoxin contamination, and recent advancements in CRISPR-based genome editing for enhancing drought tolerance and increasing plant immune responses. This will highlight key breakthroughs and future prospects for the continuing development of superior crop germplasm and cultivars and for mitigating aflatoxin contamination in food and feed supply chains. Full article

Sequential recommendation aims to model evolving user preferences based on historical interactions. Transformer-based architectures have achieved strong performance by focusing on user-level sequential patterns, yet global item–item relationships are often underrepresented, limiting the ability to capture broader contextual signals. In many real-world scenarios, items contain rich textual attributes such as descriptions and categories, but these semantic features are seldom exploited in existing sequential models. To address this gap, a Semantic-Aware Global Graph-Enhanced Sequential Recommendation framework (SAGERec) is developed, in which globally derived semantic structures are incorporated to enrich item representations before sequence modeling. Large language models (LLMs) are used to generate semantically grounded item embeddings, from which a global item–item graph is constructed to capture content-level relations that extend beyond behavioral co-occurrence. These semantic relations are further refined through an adaptive edge-weight learning mechanism, enabling the graph structure to align with evolving item representations during training. The adaptively enhanced item embeddings are subsequently integrated into a lightweight Transformer-based sequential encoder for next-item prediction. Extensive experiments on three benchmark datasets demonstrate that the proposed framework consistently outperforms competitive baselines, indicating that integrating LLM-derived semantics with adaptive graph refinement leads to more expressive sequential representations. Full article

The brain and the immune system communicate in many ways and interact directly at neuroimmune interfaces at brain borders, such as hippocampus, choroid plexus, and gateway reflexes. The first part of this review described intercellular communication (synapses, extracellular vesicles, and tunneling nanotubes) during homeostasis and neuroimmunomodulation upon dysfunction. This second part compares spatial orientation, learning, and memory function in both systems. The hippocampus, deep in the medial temporal lobes of the brain, is reported to play a central role in all three functions. Its medial entorhinal cortex contains neuronal spatial cells (place cells, head direction cells, boundary vector cells, and grid cells) that facilitate spatial navigation and allow the construction of cognitive maps. Sensory input (about 100 megabytes per second) via engram neurons and top down and bottom up information processing between the temporal lobes and other lobes of the brain are described to facilitate learning and memory function. Output impulses leave the brain via approximately 1.5 million fibers, which connect to effector organs such as muscles and glands. Spatial orientation in the immune system is described to involve gradients of chemokines, chemokine receptors, and cell adhesion molecules. These facilitate immune cell interactions with other cells and the extracellular matrix, recirculation via lymphatic organs (lymph nodes, thymus, spleen, and bone marrow), and via lymphatic fluid, blood, cerebrospinal fluid, and tissues. Learning in the immune system is summarized to include recognition of exogenous antigens from the outside world as well as endogenous blood-borne antigens, including tumor antigens. This learning process involves cognate interactions through immune synapses and the distinction between self and non-self antigens. Immune education via vaccination helps the process of development of protective immunity. Examples are presented concerning the therapeutic potential of memory T cells, in particular those derived from bone marrow. Like in the brain, memory function in the immune system is described to be facilitated by priming (imprinting), training, clonal cooperation, and an integrated perception of objects. The discussion part highlights evolutionary aspects. Full article

With the rapid growth of smart devices and positioning technologies, spatiotemporal data has become essential for predicting user behavior. However, many existing next-location prediction models employ oversimplified temporal modeling, neglect spatial structure and semantic relationships, and fail to capture complex location interaction patterns. This study proposes a graph neural network model that integrates spatiotemporal features to enhance next-location prediction. There are three components in the proposed method. The first is location feature representation which combines geocodes and location category embeddings to construct semantically enriched node representations. The second is temporal modeling which computes temporal similarity between historical trajectories and current behaviors to generate time-decay weights, thereby capturing behavioral periodicity and preference shifts. The third is preference integration which long-term historical preferences and short-term current preferences are modeled using a long short-term memory (LSTM) network and subsequently fused with spatial preferences to generate a comprehensive semantic representation encompassing both user preferences and location characteristics. Experiments on real-world trajectory datasets demonstrate that our proposed model achieves superior accuracy compared to state-of-the-art approaches in next-location prediction. Full article

This review summarizes the interactions between three major bacterial groups, Rickettsia sp., Bartonella sp. and Yersinia pestis, the flea vectors and the diverse gut microbiota of fleas and highlights open questions. The focus is on the plague pathogen, Y. pestis, which adapted to transmission by fleas several thousand years ago. This caused one of the deadliest infectious diseases known to mankind, and the three pandemics resulted in an estimated 200 million deaths. In the vector, Y. pestis resists the adverse conditions, like other numerous bacterial species. Rickettsia sp. and Bartonella sp. as well as Y. pestis induce specific changes in the microbiota. The presence of bacteria in the ingested blood activates the production of antimicrobial proteins and reactive oxygen species, which normally have no effect on the development of Y. pestis. This bacterium infects mammals by different modes, first by an early-phase transmission and then by biofilm-mediated blockage of the foregut. Both interfere with blood ingestion and lead to reflux or regurgitation of intestinal contents containing Y. pestis into the bite site. Blockage of the gut leads to more attempts to ingest blood, increasing the risk of transmission. The lifespan of the fleas is also reduced. As Y. pestis is still endemic in wildlife in many regions of the world and human infections continue to occur in limited areas, studies of the interactions are needed to find new ways to control the disease. Full article

Object detection is an important tool in many areas, such as robotics or autonomous driving. Especially in these areas, a wide variety of object classes must be detected or interacted with. Models from the field of Open-Vocabulary Detection (OVD) provide a solution here, as they can detect not only base classes but also novel object classes, i.e., those classes that were not seen during training. However, one problem with OVD models is their poor calibration, meaning that the predictions are often too over- or under-confident. To improve the calibration, Temperature Scaling is used in this study. Using YOLO World, one of the best-performing OVD models, the aim is to determine the optimal T-value for this calibration method. For this reason, it is investigated whether there is a correlation between the logit distribution and the optimal T-value and how this can be modeled. Finally, the influence of Temperature Scaling on the Expected Calibration Error (ECE) and the mAP (Mean Average Precision) will be analyzed. The results of this study show that similar logit distributions of different datasets result in the same optimal T-values. This correlation could be best modeled using Kernel Ridge Regression (KRR) and Support Vector Machine (SVM). In all cases, the ECE could be improved by Temperature Scaling without significantly reducing the mAP. Full article

Crowdsourced delivery plays a key role in fresh food retailing, where tight time limits and perishability require fast, reliable fulfillment. However, real-time order–courier assignment is challenging because orders arrive in bursts, couriers’ locations and availability change, capacities are limited, and many decisions must be made simultaneously. We propose Attn-QMIX, a novel attention-augmented multi-agent reinforcement learning framework that models each order as an agent and learns coordinated matching strategies through centralized training with decentralized execution. The framework develops a new capacity-aware multi-head attention mechanism that captures complex order–courier interactions and dynamically prevents courier overload and integrates it with a QMIX-based mixing network equipped with hypernetworks to enable effective credit assignment and global coordination. Extensive experiments on a real-world road network show that Attn-QMIX outperforms five representative methods. Compared with a novel cooperative ant colony optimization method, it reduces total cost by up to 2.30% while being up to 3403 times faster in computation. Full article

Ginger is a common spice found in many cuisines all over the world that is from the rhizome of Zingiber officinale. Additionally, it has been used in traditional medicinal practices as an aid in many ailments ranging from nausea to muscle pain. The non-volatile compounds of ginger, including zingerone, are responsible for pungency and they have widespread biomedical activities. The crystal structure of zingerone, a 6-gingerol degradation product and phenolic compound, reveals that the C4 hydroxyl group is the fulcrum for strong intermolecular interactions such as (O1-H2…O3) 2.737(2) Å. Our electrochemical results using rotating ring-disk electrode (RRDE) hydrodynamic voltammetry demonstrate that zingerone is an effective scavenger of superoxide radical anions and that zingerone, unlike powdered ginger, is a strong antioxidant with a collection efficiency slope of −6.5 × 10⁴ M⁻¹. The addition of vitamin C enhances scavenging activity for both zingerone and ginger powder, although the effect is more noticeable with zingerone. Correspondingly, the zingerone/vitamin C efficiency slope value is −5.40 × 10⁵ M⁻¹. Density Functional Theory (DFT) calculations permit the development of a plausible antioxidant mechanism for zingerone, and zingerone synergistic action with vitamin C, in which zingerone is capable of being regenerated with the assistance of protons that may be provided by ascorbic acid. This mechanism demonstrates that zingerone acts as a strong antioxidant agent by virtue of its C4 hydroxyl group and aromatic system. The scavenging chemical reaction is the same as that obtained through the dismutation of superoxide by superoxide dismutase (SOD) enzymes into hydrogen peroxide and molecular oxygen. Thus, zingerone behaves as a SOD mimic. Full article

While Artificial Intelligence (AI) promises significant enhancements for Clinical Decision Support Systems (CDSSs), the opacity of many AI models remains a major barrier to clinical adoption, primarily due to interpretability and trust challenges. Explainable AI (XAI) seeks to bridge this gap by making model reasoning understandable to clinicians, but technical XAI solutions have too often failed to address real-world clinician needs, workflow integration, and usability concerns. This study synthesizes persistent challenges in applying XAI to CDSS—including mismatched explanation methods, suboptimal interface designs, and insufficient evaluation practices—and proposes a structured, user-centered framework to guide more effective and trustworthy XAI-CDSS development. Drawing on a comprehensive literature review, we detail a three-phase framework encompassing user-centered XAI method selection, interface co-design, and iterative evaluation and refinement. We demonstrate its application through a retrospective case study analysis of a published XAI-CDSS for sepsis care. Our synthesis highlights the importance of aligning XAI with clinical workflows, supporting calibrated trust, and deploying robust evaluation methodologies that capture real-world clinician–AI interaction patterns, such as negotiation. The case analysis shows how the framework can systematically identify and address user-centric gaps, leading to better workflow integration, tailored explanations, and more usable interfaces. We conclude that achieving trustworthy and clinically useful XAI-CDSS requires a fundamentally user-centered approach; our framework offers actionable guidance for creating explainable, usable, and trusted AI systems in healthcare. Full article

In response to the sustainability challenges of mining, restrictive policies aimed at improving ecological quality have been enacted in various countries and regions. The purpose of this study is to examine the environmental changes in the Ningdong mining area, located on the Loess Plateau, over the past 25 years, due to many factors, such as coal mining, using the area as a case study. In this study, Landsat satellite images from 2000 to 2024 were used to derive the remote sensing ecological index (RSEI), while the RSEI results were comprehensively analyzed using the Sen+Mann-Kendall method with Geodetector, respectively. Simultaneously, this study utilized land use datasets to calculate the ecological grade (EG) index. The EG index was then analyzed in conjunction with the RSEI. The results show that in the time dimension, the ecological quality of the Ningdong mining area shows a non-monotonic trend of decreasing and then increasing during the 25-year period; The RSEI average reached its lowest value of 0.279 in 2011 and its highest value of 0.511 in 2022. In 2024, the RSEI was 0.428; The coupling matrix between the EG and RSEI indicates that the ecological environment within the mining area has improved. Through ecological factor-driven analysis, we found that the ecological environment quality in the study area is stably controlled by natural topography (slope) and climate (precipitation) factors, while also being disturbed by human activities. This experimental section demonstrates that ecological and environmental evolution is a complex process driven by the nonlinear synergistic interaction of natural and anthropogenic factors. The results of the study are of practical significance and provide scientific guidance for the development of coal mining and ecological environmental protection policies in other mining regions around the world. Full article