Vol. 11, No. 05 [May 2025]

This article explores current strategies for scaling frontend architectures through distributed data storage and delivery. It offers a detailed examination of modular and component-based design patterns, microfrontend architectures, server-side rendering (SSR), edge rendering, CDN caching mechanisms, and the integration of GraphQL. The discussion also covers advanced techniques for multi-level caching (including CDN, Redis, Service Workers, and IndexedDB), fault tolerance (such as autoscaling, circuit breakers, and blue-green deployments), data consistency (TTL management, Pub/Sub systems, and optimistic locking), and web security (TLS, JWT/OAuth, CSP, and Subresource Integrity). Drawing from a broad literature review, the article proposes a holistic framework for designing frontend systems that are high-performing, resilient, and secure. Findings suggest that the combination of microfrontend principles with a flexible cache invalidation strategy and distributed storage can significantly reduce latency while enhancing fault tolerance. The practical contribution lies in offering actionable architectural guidelines for developers and engineers building web and mobile interfaces capable of managing heavy traffic loads while maintaining strong data protection. This material will be particularly useful for researchers and system architects involved in the development of geographically distributed frontend applications. It sheds light on the trade-offs between consistency, availability, and latency in distributed client-side environments. Additionally, it may serve as a valuable reference for PhD candidates and postgraduates in computer science focused on the theoretical foundations of distributed systems and real-world techniques for optimizing throughput and reliability in contemporary web platforms.
Keywords: frontend architecture, microfrontends, distributed caching, CDN, server-side rendering, edge rendering, GraphQL, fault tolerance, data security.

This study describes the construction of fault-tolerant and scalable solutions based on open-source BI platforms under rapid growth in enterprise data volume and the evolution of business analytics. The relevance of the work is determined by the forecasted doubling of the global BI-platform market by 2032 and organizations’ growing need for continuous, embedded analytics with guaranteed response times, making the combination of horizontal scaling, multi-tier caching, and a fault-tolerant storage layer critical. The novelty of the research lies in the systematic integration of three directions: the evolution of Metabase’s built-in cache mechanisms, advanced Kubernetes autoscaling practices (configuring HPA/VPA by the active_query_count metric and applying GitOps patterns), and optimizations of storage engines (PostgreSQL 17, Snowflake dynamic tables). The author’s empirical case study—covering end-to-end Metabase integration for over one hundred organizations—confirms the practical effectiveness of the proposed approaches.
The main conclusions are: first, hybrid caching with differentiated TTLs by query type and predictive invalidation significantly reduces storage load without sacrificing interactivity; second, fine-tuned Kubernetes HPA/VPA based on Usage Analytics ensures stable replicas and optimal resource utilization; third, a PostgreSQL shared-schema model combined with a Snowflake offload layer enables both tenant-count scaling and strict data isolation, while reducing infrastructure costs by up to 50%.
This article will be helpful to engineering and DevOps teams responsible for building and maintaining high-load BI solutions on open-source platforms.
Keywords: Metabase, scaling, caching, Kubernetes, multi-tenancy, performance, BI platforms, open-source.

One of the longstanding concerns within society is the prevalence of infectious diseases. Research has shown that certain infections, particularly those involving multiple pathogens, complicate the identification and treatment of affected individuals, thereby adversely impacting public health. Consequently, a model for co-infection involving COVID-19 and malaria has been developed and analyzed to explore the effects of threshold quantities and the co-infection transmission rate on the synergistic interaction between the two diseases. This model enhances our understanding of the co-dynamics of these infections within the population. Initially, the existence and stability of the disease-free equilibrium for each individual infection were assessed using their respective reproduction numbers. It was determined that the COVID-19 and malaria-free equilibrium is locally asymptotically stable when the individual threshold quantities, Rc and Rm are less than one. In contrast, the transmission of malaria is influenced by the size of the vector population, which is determined by the recruitment rate Λv , alongside an increase in the effective biting rate am , the probability of the malaria transmission per mosquito bite βm , and the likelihood of transmission from infected humans to vectors βv
Keywords: Co-infection, COVID-19, Malaria, Effective reproduction number, Equilibrium points, Sensitivity analysis

Field experiment was conducted from July-November, during the 2024 planting season at the Akperan Orshi Polytechnic, Yandev, Benue State, Nigeria to evaluate the effect of N-and P- fertilizer rates that will maximize yield of upland rice–soybean intercrop under Yandev conditions. The treatments consisted of four fertilizer rates (0 kg Nha: 0 kg P/ha, 40 kg N/ha: 20 kg P/ha, 80 kg N/ha: 40 kg P/ha and 120 kg N/ha: 60 kg P/ha) replicated three times in a Randomized Complete Block Design. The implication of the study showed that the highest yield of upland rice was obtained at fertilizer rate of120 kg N/ha: 60 kg P/ha while that of soybean was 80 kg N/ha: 40 kg P/ha and could therefore be recommended for Yandev, location, Nigeria.
Keywords: Rice, Soybean, Intercrop, Fertilizer rate.

This article explores the potential of predictive financial modeling using machine learning techniques to support budgetary discipline in capital programs. The study examines traditional budget forecasting methods and highlights their limitations, underscoring the need to adopt more advanced approaches such as regression analysis, time series forecasting, neural networks, and ensemble methods. Drawing on findings from prior research, the article presents a model that includes sequential stages: data collection and preprocessing, feature selection and extraction, model training, validation, optimization, and forecast generation. Implementation of the model demonstrated a reduction in forecasting errors to 5–10%, shortened reporting preparation time, and improved budgetary control. The novelty of this work lies in the integration of modern machine learning techniques into budget forecasting systems, addressing an existing gap in the literature and proposing an interdisciplinary framework that brings together experts in finance, IT, and data analytics. The results reveal that predictive financial modeling enhances the timeliness of managerial decision-making, reduces operational risk, and improves resource allocation under conditions of high market uncertainty.
Keywords: predictive financial modeling, budget forecasting, machine learning, capital programs, budgetary discipline, regression analysis, time series, neural networks, ensemble methods.

This research presents an Enhanced Genetic Algorithm-based Gesture Recognition (E-GAGR) system, an intelligent and adaptive hand gesture recognition framework designed to overcome the limitations of traditional input methods such as keyboards, mice, and touch screens. By leveraging deep learning models, Genetic Algorithm (GA) based feature optimization, and dynamic noise reduction, the system achieves superior recognition accuracy while maintaining real-time adaptability to diverse environmental conditions. The proposed architecture integrates multi-modal sensor fusion, depth-sensing technology, and predictive gesture modelling, ensuring robustness across varying lighting conditions and user interactions. Experimental evaluations conducted with 50 participants across multiple lighting environments and interaction scenarios yielded an average recognition accuracy of 98.5%, significantly outperforming traditional machine learning approaches such as CNN (91.5%), SVM (88.2%), and MLP (89.7%). Additionally, the system achieved a mean response time of 0.75 seconds, demonstrating a 30% improvement in user efficiency compared to conventional input devices. The findings establish E-GAGR as a scalable, high-performance solution for contactless human-computer interaction, with promising applications in medical interfaces (e.g., sterile surgical environments), augmented reality (AR), virtual reality (VR), smart home automation, and assistive technologies.
Keywords: Hand gesture recognition, intelligent GUI, depth sensing technology, adaptive learning, human-computer interaction, multi-modal sensor fusion, predictive gesture modelling, augmented reality, smart automation.

Ocean acidification and hypoxia are severe threats to the marine ecosystems and, thus, establishing strong forecasting tools serves as a preventive approach. In an attempt for real-time multivariate forecasting of ocean acidification and hypoxic events in coastal regions, a hybrid layered-stack ARIMALSTM framework has been proposed. Models in the current literature range from statistical ones such as ARIMA to deep learning models such as LSTM. Herein, the models are combined to exploit the best of their capabilities with respect to the datasets, which contain historical data from GLODAP and SOCAT and real-time measurements from IoT sensors, to predict oceanographic parameters such as pH, dissolved oxygen, and dissolved inorganic carbon. The stacked ensemble approach amalgamates the prowess of ARIMA in statistical modeling with that of LSTM in capturing nonlinear behavior to boost prediction accuracy. The framework is tested on the historical dataset, proving the better performance of the proposed framework over standalone models. The results are presented on a geospatial dashboard, helping stakeholder groups track environmental trends and take interventions as necessary. The study thus contributes to sustainable marine management, neatly tying into United Nations Sustainable Development Goals (UN SDGs) 13: Climate Action and 14: Life below Water, through the provision of a scalable solution for coastal ecosystems globally.
Keywords: Ocean acidification, hypoxic events, hybrid ARIMA-LSTM, multivariate forecasting, developing nations, IoT sensors, socioeconomic impacts, deep learning, geospatial visualization, sustainable marine management.