方法对比

	新报贩模型	经济订货批量 (EOQ)	随机优化
领域	运筹学	运筹学	优化
方法族	Regression model	Regression model	Process / pipeline
起源年份	1951	1913	1951 (SGD); 2014 (Adam)
提出者	Arrow, Harris & Marschak	Ford W. Harris	—
类型	Stochastic single-period inventory optimization	Deterministic inventory optimization model	Gradient-based iterative optimization
开创性文献	Arrow, K. J., Harris, T., & Marschak, J. (1951). Optimal inventory policy. Econometrica, 19(3), 250–272.	Harris, F. W. (1913/1990). How many parts to make at once. Operations Research, 38(6), 947–950 (reprint).	Robbins, H. & Monro, S. (1951). A Stochastic Approximation Method. Annals of Mathematical Statistics, 22(3), 400-407.
别名	Newsboy Model, Single-Period Inventory Model, Christmas Tree Problem, Gazete Satıcısı Modeli	Wilson EOQ Model, Harris-Wilson Model, Optimal Lot Size Model, Ekonomik Sipariş Miktarı	Stokastik Optimizasyon (SGD & Varyantları), stochastic gradient descent, SGD, Adam
相关	3	3	3
摘要	The Newsvendor Model is a single-period stochastic inventory optimization framework that determines the profit-maximizing order quantity when demand is uncertain and unsold units cannot be carried forward. Formally introduced by Arrow, Harris, and Marschak (1951) in their foundational work on optimal inventory policy, the model balances the cost of ordering too much (overage) against the cost of ordering too little (underage) to yield a closed-form optimality condition known as the critical ratio.	The Economic Order Quantity (EOQ) is a classic deterministic inventory model that identifies the order quantity minimizing the sum of annual ordering and holding costs. Introduced by Ford W. Harris in 1913 and later popularized by R. H. Wilson, EOQ assumes constant demand, fixed cost parameters, and instantaneous replenishment. It remains the foundational benchmark for inventory management in manufacturing, retail, and supply chain contexts where demand is relatively stable and costs are well-characterized.	Stochastic optimization is a family of iterative methods that minimize an objective function by computing gradients on randomly sampled subsets of data — mini-batches — rather than on the entire dataset at once. Pioneered by Robbins and Monro in 1951 as stochastic approximation, the approach became the standard engine for training large-scale machine-learning models through variants such as SGD with momentum, AdaGrad, RMSProp, and Adam.