Browse Articles

Full blood count dynamics in immunologically naïve individuals with mild COVID-19: A prospective community cohort study

Purpose Full blood count (FBC) provides a range of cellular haematological parameters and serves as a routinely available basic immune profile. While FBC has been widely used to monitor moderate-to-severe COVID-19 in hospitalised settings, its temporal dynamics in mild, community-managed cases remain poorly characterised, despite these constituting the majority of global infections. Methods In a prospective cohort study, we tracked the cellular haematological profiles of 93 recently exposed, immunologically naïve individuals with mild COVID-19 and no underlying co-morbidities, recruited to the Integrated Network for Surveillance, Trials and Investigations into COVID-19 Transmission (INSTINCT) study. Blood samples were collected on D0, D7, D14 and D28 and subsequently aligned to infection-timepoints based on the day of first detected PCR positivity, and analysed using mixed-effects models. Results Over 30% of cases exhibited transient, clinically defined neutropenia (mean (95% CI): First PCR-Positive (FP) 2.47x10 9 /L (2.26–2.68) vs convalescence 3.34 x10 9 /L (2.98–3.7); p = 0.0013). Over 20% exhibited lymphopenia (FP 1.38x10 9 /L (1.28–1.48) vs convalescence 1.79x10 9 /L (1.79–2.01); p = 0.0013). Additionally, we observed a notable elevation in platelet count, peaking approximately two weeks after initial infection (mean (95% CI): FP + 14 283x10 9 /L (254–311) vs convalescence 237x10 9 /L (222–252); p = 0.0013). Conclusions Transient neutropenia and lymphopenia occurred in approximately one-third and one-fifth of mild COVID-19 cases, respectively, and were followed by a delayed increase in platelet count. This study provides a descriptive, prospective dataset of full blood count parameters spanning early infection to convalescence in immunologically naïve individuals with mild SARS-CoV-2 infection. These data may support mathematical modelling of within-host cellular haematological dynamics and have potential clinical relevance for understanding typical trajectories of routine blood parameters during mild disease.

Self-positioning in space science communication: A corpus-assisted discourse study

This study incorporates text mining into critical discourse analysis to examine how government science agencies in China and the United States position themselves in space science communication on social media. Two specialized corpora have been built by collecting posts from government science agencies on Weibo from China and X (formerly Twitter) from the U.S. With the help of the text mining tool KH Coder, this study gives a corpus-assisted discourse analysis of the particular ways of positioning at different levels of discourse: (1) topics/themes, (2) addressing terms, and (3) those words that co-occur with self-addressing terms in a sentence. The findings reveal significant differences in their preferential ways of positioning. Chinese government science agencies present themselves as state-affiliated yet approachable institutions, blending achievements and operational efficiency with patriotism and collective pride. Their use of diverse addressing terms and co-occurrence patterns portrays them as experienced, supportive guides, balancing national identity with interpersonal closeness. In contrast, U.S. government science agencies emphasize professionalism, focusing on research, space exploration, mission execution, and audience engagement, with minimal reference to state affiliation. Their addressing terms are formal and standardized, with pride centred on mission success and discovery, highlighting expertise and scientific leadership. Their preferential ways of positioning are further explained in their respective contexts in order to present a proper understanding of these differences.

Early life stress enhances the association between residential nature exposure and fasting blood glucose

Background Emerging epidemiological evidence indicates that groups in low socioeconomic positions exhibit more pronounced health benefits from nature exposure compared to more privileged groups. We have previously posited one possible mechanism underlying this phenomenon through our framework: (susceptibility to stress) groups in low socioeconomic positions are often exposed to more early-life stressors, which can induce a lifelong susceptibility to stress through various neurobiological pathways; (environmental sensitivity) susceptibility to stress, traditionally understood as heightened reactivity to stressors, could also encompass enhanced responsivity to health-protective exposures, inducing greater risks in adverse environments, but also greater benefits in protective environments. Objective Examine the moderation effect of early life stress on the association between residential nature exposure and fasting glucose. Methods We assessed the impact of residential nature exposure (Normalized Difference Vegetation Index) on glucose dysregulation (elevated levels of fasting blood glucose) with a specific focus on the moderation effect of early life stress (Stress and Adversity Inventory for Adults) using baseline data from a cohort of 340 nursing students. Results An initial analysis did not support our linear dose-response hypothesis. However, a theory-guided exploration revealed a significant curvilinear trend wherein participants with higher but also lower exposure to early-life stressors both exhibited lower levels of fasting glucose when living in greener neighborhoods. By contrast, for participants with relatively moderate early-life stressor exposure, there was no association between neighborhood greenness and fasting glucose. Discussion Our findings contribute to growing evidence and further support the idea that increasing access to nature within disadvantaged neighborhoods could be an effective strategy to mitigate metabolic risks and attenuate health disparities among vulnerable populations. As the evidence for this framework expands, it could inform more targeted interventions that leverage individual differences in environmental sensitivity to promote health equity, ultimately providing more nuanced and socioeconomically attuned approaches to public health.

Metataxonomic insights into lactic acid bacteria diversity in artisanal coalho cheese from the Caatinga biome

Artisanal raw-milk cheeses are complex microbial ecosystems that reflect local environments through spontaneous fermentation driven by lactic acid bacteria (LAB), processing practices, and the quality of raw materials. The Caatinga biome, a unique semi-arid ecosystem in Northeastern Brazil, presents distinctive environmental and cultural conditions which can significantly influence the microbial composition of traditional dairy products. This study aimed This study aimed to characterize the bacterial community, with emphasis on lactic acid bacteria (LAB), in artisanal Coalho cheese produced in the Seridó region using 16S rRNA gene-based metataxonomic analysis. A total of 32 cheese samples from eight municipalities were collected and homogenized into a representative composite sample. DNA was extracted and the V3–V4 region of the 16S rRNA gene was sequenced using the Illumina MiSeq platform (paired-end 2 × 300 bp). The microbial community was dominated by LAB, particularly Enterococcus (~25%), followed by Lactococcus, Streptococcus, and Leuconostoc. The predominance of these genera reflects the spontaneous fermentation process typical of artisanal raw-milk cheeses and highlights their role in acidification, flavor development, and microbial stability. The microbial community was dominated by LAB, particularly Enterococcus (~25%), followed by Lactococcus , Streptococcus , and Leuconostoc . The predominance of these genera reflects the spontaneous fermentation process typical of artisanal raw-milk cheeses and highlights their role in acidification, flavor development, and microbial stability. This study provides the first metataxonomic insight into the microbiota of artisanal Coalho cheese produced in the Seridó region of the state of Rio Grande do Norte, highlighting its microbial richness and biotechnological potential, and supporting its valorisation and the sustainable development of regional cheese production. Although the use of pooled samples limits the assessment of intra-regional variability, the results establish a foundational understanding of the microbial community structure and support future investigations focused on functional characterization, food safety, and process standardization.

Parental knowledge, vaccine hesitancy, and practices regarding seasonal influenza vaccination for preschool-aged children in Shenzhen, China: Insights from a cross-sectional survey

Background Seasonal influenza poses a substantial health threat to children under 5 years of age. Despite recommendations for annual seasonal influenza vaccination (SIV), coverage among preschool-aged children in China remains low. Vaccine hesitancy is an important barrier to vaccine uptake. This study assessed parental knowledge, vaccine hesitancy, and practices regarding SIV for preschool-aged children in Shenzhen, China, and identified factors associated with hesitancy. Methods A cross-sectional online survey was conducted from September to December 2025 using a multistage sampling strategy across six administrative districts of Shenzhen. Parents of children aged 0–5 years were recruited from community health service centers and kindergartens/childcare institutions. An adapted Chinese version of the Parental Attitudes toward Childhood Vaccines (PACV) scale was used to assess hesitancy toward SIV. Parents’ knowledge regarding influenza and SIV, as well as SIV practices, were assessed and scored. Independent factors associated with hesitancy scores were identified using multivariable linear regression, and factors associated with binary hesitancy outcomes were conducted using multivariable logistic regression. Sensitivity analyses excluding the behavioral PACV items were also performed. Results A total of 17,955 parents were included in the final analysis, and 19.16% reported that their child had received SIV in the past year. The mean PACV hesitancy score was 32.26 ± 18.50, and 19.42% of parents (3,487/17,955) were classified as vaccine hesitant. The highest proportions of hesitant responses were related to concerns about serious adverse reactions (69.47%), the safety of repeated annual vaccination (55.61%), and doubts regarding vaccine effectiveness (52.68%). Higher knowledge and vaccination practice scores were independently associated with lower levels of vaccine hesitancy in both multivariable linear and logistic regression analyses (both P  < 0.001). Additionally, older parental age, Shenzhen household registration, employment in enterprises/public institutions, middle-level household income, kindergarten attendance, obtaining vaccine information from kindergartens and relatives/friends were negatively associated with vaccine hesitancy in multivariable logistic regression analyses (all P  < 0.05). Sensitivity analyses yielded generally consistent findings. Conclusion Parental hesitancy toward SIV among preschool-aged children in Shenzhen was low-to-moderate and was associated with knowledge, prior vaccination behaviors, and sociodemographic characteristics. Concerns about vaccine safety and effectiveness were the most frequently cited reasons for hesitancy. Targeted communication strategies delivered through multiple channels, together with consideration of free SIV for preschool-aged children may help improve vaccine uptake in this high-burden population.

Prevalence, diagnostic methods, and clinical outcomes of wasting among paediatric cancer patients in Africa: A systematic review and meta-analysis

Background Malnutrition is a prevalent and modifiable co-morbidity in paediatric oncology, yet no comprehensive pan-African synthesis of its burden has been conducted. This systematic review and meta-analysis estimated the pooled prevalence of wasting and acute malnutrition among paediatric cancer patients in Africa, evaluated the impact of diagnostic assessment methods on reported prevalence, and characterised associations with adverse clinical outcomes. Methods We searched PubMed/MEDLINE, EMBASE and Web of Science for observational studies published from January 2000 to December 2025 reporting wasting or acute malnutrition among children and adolescents (0–19 years) with confirmed malignancies in African healthcare settings. Two independent reviewers screened studies, extracted data, and appraised quality using the Joanna Briggs Institute Critical Appraisal Checklist. Pooled prevalence was calculated using a random-effects model with Freeman-Tukey double arcsine transformation. Heterogeneity was assessed using the I ² statistic and Cochran’s Q test. Publication bias was evaluated with Egger’s regression test. Certainty of evidence was assessed using the GRADE framework. Results Sixteen independent cohorts comprising 2,419 paediatric cancer patients across eight African countries were included. The pooled prevalence of wasting or acute malnutrition was 39.7% (95% CI: 30.7%–49.1%), with individual study estimates ranging from 11.7% to 67.6%. Between-study heterogeneity was substantial ( I ² = 95.3%; 95% CI: 92.7%–96.1%, p < 0.001). Studies using mid-upper arm circumference (MUAC) consistently detected substantially higher rates of malnutrition than those relying on weight-for-height or body mass index criteria from the same patient populations. No significant publication bias was detected (Egger’s test p = 0.503). GRADE certainty of evidence was very low, primarily due to heterogeneity. Wasting was independently associated with higher chemotherapy toxicity, treatment-related neutropenia, sepsis, and reduced overall survival across two of the included cohorts. Conclusions Wasting and malnutrition affects about two in five children with cancer in the African countries studied. The diagnostic tool employed is the single most consequential determinant of detected prevalence, and weight-based metrics alone are inadequate in populations with large solid tumours. Universal adoption of MUAC-based screening, structured algorithm-guided nutritional intervention, and integration of socioeconomic vulnerability assessment into routine oncological care are evidence-based priorities for improving treatment tolerability and survival in this population.

Epidemiology and healthcare burden of non-fatal maxillofacial injuries in Bangladesh: Evidence from Bangladesh Health and Injury Survey (BHIS), 2016

Background Maxillofacial injuries cause significant morbidity and contribute greatly to the global healthcare burden, yet epidemiological estimates from Bangladesh are lacking. This is the first population-level study to provide nationally representative estimates of the epidemiology, healthcare burden, and economic impact of such injuries in Bangladesh. Methods A secondary analysis was conducted using data from the Bangladesh Health and Injury Survey 2016, a national cross-sectional survey that utilised multistage cluster sampling to collect data from 292,217 individuals across all age groups with a six-month recall period. Annualised incidence rates (IRs) and incidence rate ratios (IRRs), with 95% confidence intervals (CIs), were estimated and stratified by sociodemographic, injury, and healthcare-related characteristics. Results The annual IR of non-fatal maxillofacial injuries was 793 (95% CI 760.3–824.8) per 100,000 population in 2016. Males (IR 989.1 per 100,000; 95% CI 938.1–1040.2), children aged 0–4 years (IR 1577.9 per 100,000; 95% CI 1414.9–1740.9), urban residents (IR 826.1 per 100,000; 95% CI 748.1–904.2), and auto-rickshaw/bus/truck drivers (IR 1632.7 per 100,000; 95% CI 1205.0–2060.3) experienced higher rates. Falls and road traffic accidents accounted for most cases, and 56.9% of injured individuals sustained concomitant injuries. Moreover, 47.4% of the injured did not receive any treatment, while 15.1% required hospital admission. The mean inpatient out-of-pocket (OOP) expenditure was USD 470.5 [standard deviation (SD) 1497.1]. Extrapolating the BHIS 2016 estimates to the 2026 national population projected an increase in non-fatal maxillofacial injury cases from 1.27 to 1.41 million. Conclusion A high incidence rate, a substantial treatment gap, and significant OOP costs underscore critical gaps in maxillofacial injury care in Bangladesh. Alongside reinforcing the country’s infrastructure, we call for targeted policy reforms to strengthen injury prevention, treatment, rehabilitation, and financial protection.

Research on the evaluation and spatial layout of high quality development of higher education in China

As a social subsystem, the layout of higher education is related to the degree of harmony in the economy and society. Based on the panel data from the “China Statistical Yearbook”, “China Education Statistical Yearbook” and the “Education Statistical Yearbook” of China’s provincial administrative regions from 2000 to 2022, considering factors such as sustainable economic and social development capacity, regional scientific and technological cultural development level, and spatial layout of higher education, this study determines the degree of harmonic development among the three subsystems. This paper constructs an indicator system for harmonious development of China’s higher education, and uses the method of “single indicator quantification-multi indicator synthesis-multi criteria integration” (SMI-P) to determine the harmonious development index of China’s higher education. The ArcGIS spatial analysis method is used to study the spatial distribution changes of higher education harmonious development in China’s 31 provincial administrative regions. Then, the Tobit model is used to analyze the impact of various economic and social factors, scientific and cultural factors or layout of higher education factors on the harmonious development of higher education, in order to optimize the spatial layout of China’s higher education resources, balance education development, narrow regional disparities, promote regional coordination, and enhance the overall competitiveness of higher education. Research has found that: (1) The development of higher education subsystems shows a trend of convergence; (2) Over the past 23 years (from 2000 to 2023), the spatial centroid of harmonious development of higher education (HDHE) in China shifted southward; (3) There is a strong spatial interaction between 3 subsystems of harmonious development of higher education in China(economic society subsystem, technology and culture subsystem, and higher education subsystem); (4) The main factors affecting HDHE in China include the degree of regional ecological civilization, the ratio of university students to teachers and the per capita domestic patent application authorization.

Continuity or change in the transition to Islam? A diachronic assessment of agricultural production at Old Dongola, Northern Sudan (14th–18th centuries CE)

In this paper, we present an archaeobotanical study from Old Dongola, northern Sudan, examining agricultural dynamics and subsistence strategies from the Late Christian to the Islamic period. Old Dongola, established in the 5th century, served as the capital of the Kingdom of Makuria, one of the three medieval Nubian kingdoms. Following the collapse of Makuria in the 14 th century, the city underwent significant cultural and social transformations, marked by the migration of Arab tribes and the rise of new power structures. From the 14 th to the 18 th century, Old Dongola served as the seat of a local ruler who became subordinate to the Funj Sultanate in the 16 th century, a period that witnessed religious conversion, migration, and political reorganisation associated with the gradual conversion to Islam in the region. This study is distinguished by a systematic sampling strategy, with sediments collected from every excavated context in the citadel of Old Dongola. The archaeobotanical results, spanning the Kingdom of Dongola and Funj periods (14 th –18 th centuries), shed light on the dietary habits of the site's inhabitants. The evidence indicates an increasing reliance on sorghum, alongside a gradual decline of bread wheat and hulled barley. The introduction of crops such as tobacco and eggplant from the 16 th century onwards may reflect broader economic and cultural shifts occurring during this period of religious and demographic change, though a direct causal link between these crop introductions and processes of demographic change or conversion to Islam cannot be established from archaeobotanical evidence alone. This study contributes to our understanding of agricultural change in early modern Sudan, adding to an understudied archaeobotanical record at a time of significant socio-political transition.

Evaluating a sexual violence primary prevention program in Australian secondary schools: A protocol for a pilot cluster randomised controlled trial

Background Sexual violence is a global human rights issue and a significant public health concern. Prevention of sexual violence requires addressing the structural and social norms that perpetuate it. Schools are promising settings for primary prevention interventions, as early socialisation experiences can shape attitudes and behaviours that enable sexual violence. However, evidence on effective school-based interventions is limited. The objective of this pilot study is to assess the feasibility and preliminary effectiveness of an Australian sexual violence primary prevention program, the Schools Education Program, on student sexual violence knowledge, attitudes and behavioural intentions. Methods/design A two-arm, parallel pilot cluster randomised controlled trial will be conducted with Year 9 students (aged 13–15) in 12 secondary schools across one Australian state. Ten schools will be randomly allocated to the SEP intervention (n = 6) or a waitlist control (n = 4); an additional two schools are pre-assigned to the waitlist control group. The SEP comprises three student education modules, parent and staff education, and the recruitment of School Student Champions. The primary outcome is sexual violence knowledge. Secondary outcomes include attitudes, behavioural intentions, and implementation factors. Quantitative data will be collected at baseline, immediately post-intervention, and at 6-month follow-up. Analysis will use linear mixed-effects models to provide preliminary estimates of effect and estimate key parameters for a future definitive trial. The trial is embedded within a concurrent realist evaluation that includes qualitative methods to explore the mechanisms and contextual conditions shaping these outcomes. Discussion This study is the first pilot trial to evaluate a sexual violence primary prevention intervention in Australian secondary schools. In addition to the trial focused on sexual violence prevention outcomes, its integration with realist evaluation methodology will advance knowledge about how, when, and for whom these outcomes occur. The study findings will inform adaptability and scalability for secondary schools nationally and internationally. Trial registration The trial is prospectively registered with the Australian and New Zealand Clinical Trials Registry (reference: ACTRN12625000409471p). Registered 6 May 2025. Protocol version v.1.0, 20/05/2025

Improving the care of people affected by post-COVID syndrome (LCovB):study protocol of a mixed-methods study

Background The COVID-19 pandemic has resulted in a large number of individuals affected by post-COVID syndrome (PCS) and poses significant challenges to healthcare provision. Many PCS patients are left to manage their condition without adequate support. The aim of the LCovB study is to collect data on symptoms (including pre-existing conditions and comorbidities) and the care situation and trajectories of PCS patients. The results will serve to identify key measures for optimizing care. Methods Using a mixed-methods approach, the study includes secondary data analyses of routine data of PCS patients from 44 German statutory health insurance companies, a primary survey and semi-structured interviews with selected PCS patients. In addition, the situation of severely ill patients will be assessed through a physician’s visit and questionnaires. Finally, an expert panel with stakeholders from the healthcare sector will examine the study results using the Delphi method to derive recommendations for treatment and optimized care pathways. Discussion The LCovB study examines care trajectories to identify gaps and barriers for patients with PCS, aiming to develop evidence-based recommendations and guidelines for care and disease management, supported by an expert panel. Trial registration In August 2025, the study was prospectively registered in the German Clinical Trials Register: Module 1 (DRKS00037341), Module 2 (DRKS00037343), Module 3 (DRKS00037344), Module 4 (DRKS00037345).

A blood gas parameter–based assessment model for predicting poor prognosis in sepsis: A retrospective analysis of the MIMIC-IV and eICU-CRD

Background Blood gas parameters are associated with sepsis prognosis. This study aimed to develop an assessment model based on blood gas parameters for predicting patient outcomes. Methods Data were retrospectively extracted from the Medical Information Mart for Intensive Care IV (MIMIC-IV) and electronic Intensive Care Unit Collaborative Research Database (eICU-CRD). A sepsis assessment model was developed using the MIMIC-IV cohort, followed by internal validation in patients with septic shock from the same database and external validation in patients with sepsis from the eICU-CRD. Bioinformatics and machine learning clarified the relationship between the model and the primary outcome of 28-day mortality in patients with sepsis and septic shock. Results The sepsis assessment blood gas 3 (SABG-3), an assessment model incorporating PO 2 , base excess (BE), and lactate, was developed and validated as an independent predictor of 28-day mortality in patients with MIMIC-IV sepsis (odds ratio: 1.559; 95% confidence interval: 1.464–1.659; P  < 0.001). Its prognostic performance was internally validated in patients with MIMIC-IV sepsis and externally validated in patients with eICU-CRD sepsis. High-risk patients identified by SABG-3 exhibited greater illness severity than low-risk ones. Sensitivity analyses across five methods confirmed the prognostic value of SABG-3 for intensive care unit patients with sepsis and septic shock. A SABG-3-derived nomogram proved superior to existing scales. Conclusion We developed and validated a novel assessment model and nomogram to evaluate sepsis prognosis rapidly and to identify patients who may benefit from intensified treatment.

Save Hubble: the race to preserve the space telescope kicks off

Assessing the appropriateness of helicopter emergency medical services for non-traumatic emergencies in a medically underserved rural area, Japan

Background Helicopter emergency medical services (HEMS) are essential in trauma care, but their role in non-traumatic emergencies remains unclear, particularly in medically underserved rural areas where long total prehospital times and limited specialty care access complicate emergency response. Objectives To evaluate the appropriateness of HEMS dispatches in non-trauma emergencies by assessing undertriage and overtriage based on clinical necessity and total prehospital time in rural western Kochi Prefecture, Japan. Methods We conducted a retrospective cohort study of adult patients (≥18 years) requiring emergency transport for moderate to severe non-traumatic conditions during HEMS operational hours from 2017 to 2021. HEMS necessity was classified as “absolute” (treatment unavailable at local core hospitals) or “relative” (critical condition benefiting from shorter transport). Appropriateness was defined as concordance between actual transport modality and predefined necessity criteria, with undertriage and overtriage determined by comparing the actual primary HEMS dispatch requests against these criteria and regional total prehospital times. Results Of the 2,312 eligible patients, 63 received primary HEMS dispatches and 2,249 ground EMS (GEMS). Among HEMS dispatches, 56 of 63 (88.9%) were overtriaged, while seven (11.1%) were appropriately triaged. Among GEMS transports, 248 of 2,249 (11.0%) retrospectively met HEMS criteria, yielding an 11.0% undertriage rate. Overtriaged cases included patients lacking relative necessity criteria or total prehospital time advantage. Conclusions In medically underserved rural regions, clinical severity alone may be insufficient for determining HEMS necessity in non-traumatic emergencies. Incorporating total prehospital time advantage and regional infrastructure consideration may refine dispatch decision-making and improve HEMS utilization appropriateness.

Keep the Hubble and James Webb Space Telescopes alive — the science is worth the price tag

Bound states in doped charge transfer insulators

Determinants of intention to use generative AI fitness assistants: Integrating the Theory of Planned Behavior, second-order information system quality, and perceived value

As generative artificial intelligence (GAI) becomes embedded in sport and health management, GAI fitness assistants can generate and iteratively refine training plans through conversational interactions; however, their adoption remains shaped by users’ trade-offs between perceived quality and risk. Drawing on the Information Systems Success Model and the Theory of Planned Behavior (TPB), and incorporating perceived value, this study analyzed cross-sectional, self-reported data from 342 valid questionnaires using structural equation modeling to examine the relationships among second-order information system quality, subjective norms, attitude, perceived behavioral control, and category-level behavioral intention to use GAI fitness assistants. The results indicate that information system quality was significantly associated with more favorable attitude (β = 0.623) and perceived behavioral control (β = 0.485), with system quality contributing the greatest weight (0.805). Bootstrapped indirect-effect tests showed that subjective norm was significantly associated with behavioral intention through attitude, perceived behavioral control, perceived value, and the serial attitude-to-value pathway, whereas its direct effect was not significant. Perceived value is the strongest predictor of intention (β = 0.447), and attitude (β = 0.275) and perceived behavioral control (β = 0.163) are also important antecedents. Overall, the model explains 50.9% of the variance in intention to use. These findings suggest theoretical and practical implications for improving the stability of GAI fitness assistants, the reliability of their recommendations, and the communication of user value.

Say hello to hard helium

Rapid quasi-periodic reconfiguration of the accretion column in pulsar 1A 0535+262

A Multitask Learning framework for E‐commerce time series analytics: Operational risk, demand forecasting, and anomaly detection

The study presents MT-PyraRisk, a multi-task learning framework that integrates pyramidal attention mechanisms for cross-border e-commerce risk prediction. The model processes multimodal time-series data through a shared feature extraction layer and task-specific prediction modules, supporting both regression tasks, such as operational risk and demand forecasting, and classification tasks, such as anomaly detection. Experiments were conducted on the Global E-commerce Dataset and the Kaggle Network Traffic Dataset using a sliding-window time-series setting, with the data divided into training, validation, and testing sets at a ratio of 70%, 15%, and 15%, respectively. The model was evaluated using MSE and MAE for regression tasks and Precision, Recall, and F1-score for classification and anomaly detection tasks. Compared with the strong baseline Timesformer, MT-PyraRisk reduced MSE by approximately 4.3% and improved F1-score and Recall by approximately 1.1% and 2.3%, respectively, on the Global E-commerce Dataset. On the Kaggle Network Traffic Dataset, MT-PyraRisk reduced MSE by approximately 5.1% and improved F1-score, Precision, and Recall by approximately 3.3%, 2.2%, and 4.5%, respectively. Ablation results further demonstrate the contribution of the pyramidal attention mechanism, task-specific prediction layers, and joint optimization module, as removing these components led to clear performance degradation. These results indicate that MT-PyraRisk can provide an effective multi-task modeling solution for operational risk forecasting, demand prediction, and anomaly detection in cross-border e-commerce risk management.