Abstract:Island fisheries cooperatives depend on marketplaces where connectivity is intermittent and transaction logging is often delayed. We design an offline first marketplace platform that supports inventory tracking, cooperative pricing, and deferred synchronization between mobile devices and district hubs. Conflict resolution rules prioritize transaction integrity and transparent audit trails during delayed merges. Pilot deployment data from coastal cooperatives indicate better stock visibility, fewer duplicate sales records, and shorter payment reconciliation cycles. The study highlights governance requirements for community administrators and provides implementation guidance for resilient digital trade infrastructure in archipelagic economies.
Abstract:Regional archives frequently preserve historical manuscripts as fragmented or low quality scans, limiting scholarship and public access. This paper proposes a reconstruction pipeline that joins texture aware inpainting with structural language priors to restore damaged sections while preserving editorial traceability. The system assigns confidence scores to each reconstructed segment and stores provenance metadata for curator review. Case studies on multilingual manuscript collections show improved readability and downstream indexing quality without replacing expert verification. Results support a collaborative workflow where computational restoration accelerates heritage preservation while maintaining transparent boundaries between observed and reconstructed content.
Abstract:Many medium sized cities operate sparse air quality sensor networks that leave substantial gaps in local exposure estimates. We introduce a data fusion forecasting approach that combines sparse station readings, traffic proxies, and satellite aerosol indicators to produce neighborhood level forecasts up to forty eight hours ahead. The model uses uncertainty aware temporal convolution and transfer learning from adjacent regions with similar climate patterns. Evaluation across dry and wet seasons indicates improved forecast stability relative to autoregressive baselines, particularly during abrupt pollution spikes. The framework provides actionable evidence for health advisories and low cost deployment strategies in data limited urban systems.
Abstract:Rural microgrids in river basin communities face significant uncertainty from seasonal flow variation and cloud cover. This work proposes a stochastic dispatch optimizer integrating short horizon hydrological predictions with photovoltaic uncertainty envelopes. A scenario reduction strategy enables near real time scheduling while preserving risk sensitivity for battery reserves and critical loads. Comparative tests on two community microgrids show reduced curtailment and lower diesel backup use during volatile weather weeks. Sensitivity analysis reveals that uncertainty calibration frequency strongly affects reliability outcomes, suggesting operational guidelines for utilities deploying mixed renewable portfolios in resource constrained regions.
Abstract:Educational institutions running bilingual programs often assess participation through manual observation, which is time intensive and inconsistent. We present a multimodal analytics system that fuses speech segments and gesture trajectories to estimate engagement, turn taking balance, and topic continuity. The framework includes privacy aware aggregation and language adaptive transcription tuned for mixed classroom discourse. Experiments in undergraduate communication courses indicate improved agreement with instructor evaluations compared with single modality baselines. The model surfaces practical indicators for lesson redesign, including silent intervals and asymmetric participation clusters, while maintaining feasible processing costs for mid tier campus infrastructure.
Abstract:Cooperative lenders serving smallholder producers require interpretable credit decisions that align with local regulations and member trust. We design a transparent scoring pipeline that combines monotonic gradient boosting with causal feature screening to separate predictive signals from spurious socioeconomic correlations. Using five years of cooperative lending records, the model improves default discrimination while preserving explanation consistency across regional groups. Loan officers reported greater confidence when counterfactual explanations identified actionable factors such as repayment timing and inventory turnover. The study demonstrates that fairness aware and explainable scoring can improve inclusion without sacrificing portfolio stability in constrained rural credit environments.
Abstract:Flood response teams in delta settlements often lack rapid digital maps immediately after severe rainfall. This paper evaluates lightweight vision architectures trained on mixed drone and street level imagery to classify structural damage categories within hours after events. A compact encoder with knowledge distillation is proposed to run on modest municipal hardware and volunteer laptops. Results across three seasonal datasets show competitive segmentation quality with substantially lower inference time and memory demand. Error analysis highlights roof texture confusion in densely packed neighborhoods, leading to recommendations on data augmentation and local calibration protocols for emergency operations.
Abstract:This study introduces an adaptive scheduling framework for telehealth internet of things devices operating in mountainous districts where latency and packet loss vary across villages. A hybrid heuristic combines queue awareness, device battery state, and clinic urgency levels to prioritize medical signals and periodic diagnostics. Field simulations based on community clinic traces indicate lower response delay for high risk alerts and improved continuity for routine monitoring. The method also reduces dropped packets during intermittent connectivity windows by reallocating low priority workloads to resilient periods. Findings support practical deployment in constrained regional health networks.
Abstract:Emergency medical responders require high-fidelity training for rare critical scenarios. We compared virtual reality simulation with traditional mannequin-based training for paramedic students in Poland. VR groups demonstrated equivalent procedural skill acquisition with superior decision-making in complex multi-patient scenarios. Training retention at six months showed 20% better performance maintenance for VR-trained cohorts. Motion sickness affected 8% of participants requiring accommodation. Cost-effectiveness analysis favored VR at scales above 50 trainees annually. Our findings support VR integration in emergency medical education curricula.
Abstract:Intensive aquaculture antibiotic use contributes to resistance gene proliferation. We surveyed antimicrobial resistance patterns across Vietnamese shrimp and fish farms examining water, sediment, and animal samples. Multi-drug resistant bacteria prevalence reached 45% with resistance genes detected in downstream water bodies. Farm-level antibiotic use intensity correlated with resistance prevalence. Probiotic alternatives showed promise reducing therapeutic antibiotic requirements. Our surveillance data supports antimicrobial stewardship program development for aquaculture sectors.