inventory risk modeling

The newsvendor problem faces challenges in dynamic inventory forecasting due to scarce historical data and unknown demand distributions. Method: This paper proposes a distribution-free stochastic modeling framework that bypasses prior distributional assumptions. Leveraging stochastic forecasting analysis, it directly learns the evolution dynamics of inventory states from limited time-series inventory and sales data, enabling dynamic probabilistic characterization of inventory levels. Contribution/Results: Unlike conventional approaches relying on strong parametric assumptions (e.g., normal or Poisson demand), our method establishes a data-driven, distribution-agnostic dynamic modeling paradigm. Experiments on real-world e-marketplace data demonstrate that the model significantly outperforms classical distribution-based methods in short-term forecasting—achieving superior accuracy, robustness, and practical deployability. It provides an interpretable, probability-based solution for inventory decision-making under small-sample regimes.

This study investigates the sequential interaction between pricing and inventory decisions in digital retail competition, focusing on a price-then-inventory setting where demand uncertainty and strategic uncertainty induce behavioral biases. Using a combination of game-theoretic modeling and controlled laboratory experiments, it tests theoretical predictions against observed human behavior. Results reveal three key deviations: (1) retailers’ pricing decisions exhibit strong reference-price dependence while neglecting demand volatility; (2) inventory choices display systematic “pull-to-center” bias; and (3) pricing and inventory decisions are severely decoupled, with markedly lower sensitivity to profit margins and demand uncertainty than predicted by equilibrium theory. This work is the first to systematically identify and quantify these two critical behavioral biases—reference-price anchoring and pull-to-center—in a sequential operations game. It demonstrates that such biases substantially distort competitive equilibria, offering novel empirical evidence and theoretical refinements for digital platform governance and retailer operational optimization.

This study addresses the computational challenges in analyzing non-Markovian (s, S) inventory systems under general inter-demand and lead-time distributions, where analytical solutions are intractable and conventional simulation methods incur high computational costs. To overcome this, the authors propose a supervised learning–based neural network framework that leverages only low-order moments of the demand and lead-time distributions as inputs to efficiently predict key steady-state performance metrics—such as the inventory level distribution, expected cycle time, and stockout probability. The approach achieves high-accuracy, near-instantaneous predictions across a broad range of parameter settings, substantially reducing computational overhead while demonstrating strong potential for extension to other complex inventory models.

This work addresses the challenge faced by graduate students in finance and economics—whose programming backgrounds vary widely—in mastering quantitative methods. To bridge the gap between theory and practice, the project develops a unified Python toolkit that integrates probability theory, statistics, numerical methods, and empirical modeling. Through accessible, reproducible examples and exercises centered on core financial applications such as asset pricing, risk measurement, and forecasting, the toolkit lowers entry barriers while emphasizing pedagogical clarity. Key technical components include Monte Carlo simulation, numerical optimization, root-finding algorithms, and time series modeling, all implemented with a focus on vectorization, numerical stability, and interpretability of results. The resulting materials provide a systematic, transparent, and reproducible foundation for both teaching and research in quantitative finance.

This study uncovers the network-topological origins of the bullwhip effect, addressing demand uncertainty and risk cascades in supply chains. Building upon stochastic networks and the newsvendor model, it develops a quantitative framework for analyzing multistage supply chains under both equilibrium and transient dynamics. The framework integrates nonlinear price elasticity and dynamic rebalancing of trade relationships to characterize system degradation driven jointly by decentralized decision-making and physical bottlenecks. Innovatively, the bullwhip effect is attributed to intrinsic network structure, market-clearing feedback is endogenized, and stockouts are modeled as a multidimensional Skorokhod reflection problem. Empirical analysis of global oil trade data demonstrates that localized chokepoint disruptions—such as those at the Strait of Hormuz—can trigger nonlinear cascading shortages and induce systemic reallocation of national reserves.

This study addresses the critical limitation of traditional supply chain risk models, which neglect interdependencies among nodes and thereby systematically underestimate cascading failure risks. The authors propose the first quantum-native framework that maps a four-tier, 40-node supply chain onto a 40-qubit Ising Hamiltonian. By integrating ADAPT-VQE with gradient-based operator screening and density-of-states quantum phase estimation, the framework enables simulation of correlated shock propagation, real-time ranking of intervention policies, and quantification of tail risks. A key innovation lies in leveraging quantum entanglement to model cascading failures, yielding a quantum risk measure compatible with Value-at-Risk (VaR). Demonstrated at industrial scale, the approach exhibits exponential computational advantages over classical Monte Carlo methods.

This study addresses the lack of a systematic framework for identifying critical input variables and conducting sensitivity analysis under uncertainty in complex simulations, particularly in military decision-making contexts. The authors propose a unified sensitivity analysis framework that integrates local and global methods—including variance-based, derivative-based, screening, and uncertainty quantification techniques—and strategically maps these approaches to specific decision objectives such as factor prioritization, fixing, variance reduction, and mapping. Innovatively, the framework introduces a “sensitivity audit” mechanism to enhance traceability of model assumptions and promote responsible model usage. By providing a structured guide for high-dimensional, complex simulation systems, this work significantly improves model interpretability, transparency, and the credibility of decisions derived from such models.

This study addresses the lack of modular open-source tools and standardized benchmarks for systematically analyzing the bullwhip effect in multi-echelon supply chains under asymmetric costs and realistic demand patterns. To bridge this gap, we propose the first open-source Python platform integrating a scalable simulation engine with a registry-driven benchmarking framework. The platform supports plug-in modules for demand generation, ordering policies, and cost functions, while unifying six bullwhip metrics and incorporating real-world datasets (e.g., AR(1), WSTS). Leveraging abstract base classes and vectorized Monte Carlo simulation, it achieves high computational efficiency—executing 20.8 million simulations in just 7 seconds—and reveals a cumulative amplification factor of 427× in a four-tier semiconductor supply chain. Our analysis uncovers stochastic filtering upstream and demonstrates that no single metric suffices to evaluate policy performance comprehensively: the bullwhip intensity of the Order-Up-To policy varies by up to 155× between synthetic and real demand data.

This study addresses the insufficient supply chain resilience in existing research, which often overlooks inter-tier disruption dependencies. To bridge this gap, we propose a two-stage mixed-integer programming model that explicitly captures disruption dependencies among facilities and jointly optimizes resilience (via backup reallocation), agility (through mobile facilities), and carbon emission constraints. To handle service-level probability requirements, we introduce linear cuts derived from machine learning classifiers—such as L1-regularized logistic regression—as surrogates for chance constraints. The model is efficiently solved by integrating sample average approximation with a Fix-and-Relax heuristic. Computational experiments demonstrate that our approach significantly improves solution efficiency while maintaining a 95% service level, enabling rapid generation of high-quality solutions for medium- to large-scale instances.