Traditional inventory models often assume continuous or periodic replenishment, overlooking the irregular supply patterns observed in many real-world systems. This study develops a stochastic inventory model that combines deterministic inflow, Brownian fluctuations, compound Poisson replenishment events, and proportional depletion within a unified Lévy-process framework. Analytical expressions are derived for long-run moments and stationary distributions under both finite-capacity and unbounded-capacity conditions. The results demonstrate how replenishment frequency, shipment size, volatility, depletion intensity, and storage capacity influence inventory behavior and long-run variability. Monte Carlo simulations support the analytical findings. By linking stochastic storage theory with inventory management, the study provides a tractable framework for analyzing inventory systems exposed to irregular supply and offers insights relevant to inventory planning, capacity management, and decision-making under uncertainty.
Small and medium enterprises (SMEs) are pivotal drivers of employment, Gross Domestic Product (GDP) growth, and innovation globally, yet they continue to face persistent barriers to adopting data-driven technologies, and literature lacks actionable, resource-sensitive guidance for practitioners. The systematic review maps algorithm-specific applications across industries including XGBoost for credit risk, LightGBM for real-time BI automation, and Support Vector Machines for high-dimensional small datasets while identifying persistent gaps in methodological transparency, geographic diversity, and prescriptive guidance. Addressing these gaps, the paper makes three original contributions: (1) a critical synthesis distinguishing access gaps from outcome gaps in SME BI adoption; (2) an analysis of why cost-efficiency remains under-weighted relative to strategic growth as an adoption driver, despite resource constraints; and (3) a phased three-stage implementation roadmap guiding SMEs from frugal descriptive dashboards through predictive modeling to prescriptive decision automation.
Preprocessing strategy selection in business analytics typically relies on convention rather than systematic evidence, despite consuming 60-80% of project effort. This study introduces REPROPREP (v1.0), a methodological framework for validating preprocessing effectiveness assumptions through statistical analysis and cost-benefit assessment. The framework applies Benjamini-Hochberg false discovery rate correction, quality degradation protocols, and cost-effectiveness evaluation. A demonstration across 10 UCI datasets, three preprocessing strategies, and gradient boosting classifiers with 5-fold stratified cross-validation yielded no statistically significant performance differences after multiple comparisons correction (mean effect size: 0.001 AUC), with implementation cost differences ranging from $150-$800. Focused on numeric preprocessing, REPROPREP provides organizations with a rigorous, context-specific methodology for evaluating preprocessing assumptions. Generalizability requires validation beyond tested conditions. Reproducible code is publicly available.
Preprocessing strategy selection in business analytics typically relies on convention rather than systematic evidence, despite consuming 60–80% of project effort. This study introduces REPROPREP (v1.0), a methodological framework for validating preprocessing effectiveness assumptions through statistical analysis and cost-benefit assessment. The framework applies Benjamini-Hochberg false discovery rate correction, quality degradation protocols, and cost-effectiveness evaluation. A demonstration across 10 UCI datasets, three preprocessing strategies, and gradient boosting classifiers with 5-fold stratified cross-validation yielded no statistically significant performance differences after multiple comparisons correction (mean effect size: 0.001 AUC), with implementation cost differences ranging from $150–$800. Focused on numeric preprocessing, REPROPREP provides organizations with a rigorous, context-specific methodology for evaluating preprocessing assumptions. Generalizability requires validation beyond tested conditions. Reproducible code is publicly available.
