基于马尔可夫决策的钢铁产成品订单分配模型研究  

Research on order allocation model of steel productsbased on Markov decision

作　　者：薄胜 李媛 刘海伦[3] BO Sheng;LI Yuan;LIU Hailun(Beijing Lezhi Technology Co.,Ltd.,Beijing 100089,China;China Telecom Co.,Ltd.Hebei Xiong′an New Area Branch,Xiong′an Hebei 070001,China;Beijing University of Posts and Telecommunication,Beijing 100876,China)

机构地区：[1]北京乐智科技有限公司,北京100089 [2]中国电信集团有限公司河北雄安新区分公司,河北雄安070001 [3]北京邮电大学,北京100876

出　　处：《河北省科学院学报》2024年第2期29-36,共8页Journal of The Hebei Academy of Sciences

摘　　要：物流订单分配是钢铁产成品销售过程中的重要环节,对销售过程的整体体验和全流程的良性循环起着至关重要的作用。在实际生产过程中,传统的人工分单模式已难以适应长时间维度的发展要求。为了充分考虑一个时间周期内销售的物流成本最低以及承运司机的收益较高,同时确保企业的长期收益,本文用多目标约束,建立马尔可夫决策模型,引入KM算法执行二分图匹配,基于司机收益最大化和企业成本最小的多目标优化,最大化钢铁商品总交易量为长期的最终目标,结合价值函数和多属性的多目标优化函数,形成车辆和货物的完整匹配决策。以钢铁企业的真实业务数据为例,对数据进行预处理后筛选合适特征进行模型训练并验证算法的正确性和可用性。结果表明,该模型相较于传统的订单分配方法,可以更好地解决钢铁企业在订单分配场景下的需求。Logistics order distribution is an important part of the sales process of finished steel products,and plays a vital role in the overall experience of the sales process and the virtuous cycle of the whole process.In the actual production process,long-term maintenance of manual order distribution model seems to have a richer accumulation of experience and knowledge,but in reality it has become difficult to adapt to the development requirements of a long time dimension.In order to fully consider the lowest logistics cost of sales in a time cycle as well as the higher revenue of the carrier driver,while ensuring the long-term revenue of the enterprise,the Markov decision model is established by considering multi-objective constraints,and the KM algorithm is introduced to perform bipartite graph matching.Based on the multi-objective optimization for maximizing driver revenue and minimizing enterprise cost,with the long-term ultimate goal of maximizing total steel commodity transaction volume,and combined with the value function and the multi-attribute multi-objective optimization function,a complete matching decision of vehicles and goods is formed.Using real business data of steel enterprises as an example,the data is pre-processed and then suitable features are screened for model training and the correctness and usability of the algorithm is verified.The results show that the model can better address the needs of steel companies in order allocation scenarios compared to traditional order allocation methods.

关 键 词：订单分配 马尔可夫决策 多目标优化 二分图匹配 

分 类 号：O221[理学—运筹学与控制论]