无统计信息假设下的多阶段报童决策

发布时间：2018-02-21 05:31

本文关键词： 多阶段报童问题回收价值无统计假设弱集成算法收益损失函数　出处：《中国管理科学》2015年05期 　论文类型：期刊论文

【摘要】：在不对需求做任何统计假设的情形下,该文用理论计算科学兴起的集成专家意见的弱集成算法研究多阶段报童决策。弱集成算法是一种指数加权平均集成方法,在一定的初始权重下,根据损失函数在线调整专家意见的权重。基于收益损失函数和固定订购量的专家意见,得到了与从累积收益角度研究相一致的决策方法;并扩展研究了带有回收价值的情形。理论上证明了决策方法的累积收益损失几乎不超过最优专家意见的累积收益损失。通过数值算例验证了决策方法的可行性和合理性,探讨了卖出价和成本价等因素对竞争性能的影响,说明了回收价值的引入大大提高了决策方法的竞争性能,具有重要的现实意义。
[Abstract]:Without making any statistical assumptions on demand, this paper studies multi-stage newsboy decision making by using the weak integration algorithm which integrates expert opinion, which is a kind of exponential weighted average method, which is developed by theoretical computation. Under a certain initial weight, the weight of expert opinion is adjusted online according to the loss function. Based on the expert opinion of income loss function and fixed order quantity, a decision method consistent with the study from the angle of cumulative income is obtained. It is proved theoretically that the cumulative profit loss of the decision method is almost no more than that of the optimal expert opinion. The feasibility and rationality of the decision method are verified by numerical examples. This paper discusses the influence of selling price and cost price on the competitive performance, which shows that the introduction of recovery value greatly improves the competitive performance of the decision-making method, which is of great practical significance.
【作者单位】：广东工业大学管理学院;华南理工大学工商管理学院;
【基金】：教育部人文社会科学研究项目(13YJC630234,11YJC630225) 广东省高等学校优秀青年教师培养计划(Yq2013062,Yq2013060) 国家自然科学基金资助项目(71471065,71301029)
【分类号】：F274;F224

【参考文献】