Newsvendor model variable selection method detailed in arXiv paper

Researchers have developed a new approach to variable selection for feature-based newsvendor models, which use observable covariates to optimize inventory decisions. The proposed method addresses the challenge of high-dimensional feature sets by formulating an \(\\ell_0\)-constrained empirical newsvendor problem with \(\\ell_2\)-regularization. This paper introduces a mixed-integer second-order cone programming reformulation for exact optimization and a randomized-rounding algorithm with a bi-criteria guarantee, alongside a greedy heuristic for scalability. Theoretical analysis of the sparse policy estimator includes finite-sample estimation error and out-of-sample risk bounds, with experiments on synthetic and real data demonstrating competitive operational costs and reduced covariate usage. AI

IMPACT Introduces a novel statistical method for optimizing inventory decisions using machine learning techniques.

RANK_REASON Academic paper detailing a new statistical method for a specific type of optimization problem. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv stat.ML →

• arXiv
• Big Machine Records
• Newsvendor model

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →