教育部信息管理与信息系统虚拟教研室系列活动
讲座时间：2026年10月20日15:30-17:00
讲座地点：管理楼G506
讲座主题：When Buying Ahead Reveals Demand: Procurement and Learning under Censored Demand
讲座嘉宾简介
刘鹏，新加坡国立大学统计与数据科学博士，新加坡管理大学李光前商女性自慰定量金融（实践）助理教授，兼任新加坡国立大学运筹学与分析研究所研究员。他在银行、博彩和科技等多个行业拥有超过十年的产业经验。其研究主要聚焦于量化金融、深度学习以及贝叶斯优化等领域。
讲座摘要
Managers buying storable inputs must balance price savings, holding costs, and service needs. Unrecorded shortages compound this challenge: sales conceal demand that could not be served, leaving managers unsure whether larger purchases would pay. We develop a procurement framework combining a stock target with a buy-or-wait rule. The target determines how much to buy, accounting for current service and inventory's future use; the rule compares that purchase with waiting. Executing the target also determines which demand levels sales reveal. We characterize when expected price savings above holding cost make replenishing to capacity optimal for every admissible demand estimate. These purchases then serve two purposes: capture operating savings and supply information for future decisions. We also provide a finite-sample bound on the number of informative purchases needed to resolve later ordering choices. Simulations show that routine procurement can sustain learning; when additional learning is needed, occasional purchases made to learn can cost less than robust protection. Exploratory replays that add storage to measured demand from 97 Austin buildings and 264 Portuguese electricity clients show cost reductions of 1.7% and 2.4%, respectively, from valuing carried inventory. Savings remain in retrospective comparisons at equal service. The framework helps managers capture inventory's operating value and assess whether ordinary purchases provide enough information before committing additional spending to learning.
