港口集装箱装卸作业的Q-learning动态定价策略研究

THE STRATEGY OF DYNAMIC PRICING OF HANDLING OPERATION FOR PORT CONTAINER BASED ON Q-LEARNING

港口企业在复杂的竞争环境中,需要应对不同船公司的需求制定合理的动态定价策略。使用滑动窗口方法,根据船公司减少作业时间的要求,计算出单船装卸作业时间的预期估计误差。运用TDABC(time-driven activity-based costing)方法分析码头作业成本,得出单船装卸作业的总成本。采用Q-learning算法得出以最大化单船装卸利润为主要目标,不同箱型下的动态定价策略。基于上海港实际数据分析表明:折扣因子从0. 1增加至0. 7时单船装卸利润增加至352. 77万元,从0. 7至0. 9时降至159. 89万元。相同学习率下,当单船装卸作业时间从10. 5小时减少至9小时,单船装卸利润波动明显。算例分析表明,该动态定价策略有助于港口企业的经济发展,能有效提高其竞争力。

In a complex competitive environment, port companies need to formulate reasonable dynamic pricing strategies to respond to the demands of different shipping companies. Based on the requirements of the shipping company to reduce the handling time, sliding window method was used to calculate the expected estimation error of single-vessel handling time. The TDABC （time-driven activity-based costing） method was adopted to analyze the operation cost of port and to obtain the total cost of single vessel handling operation. We used Q-learning algorithm to obtain the dynamic pricing strategy of the different box types with the main objective of maximizing single-vessel handling profits. In view of the actual data of Shanghai Port, it is concluded that when the discount factor increases from 0.1 to 0.7, the single-vessel handling profits increase to 3 527 700 yuan and when the discount factor increases from 0.7 to 0.9, the single-vessel handling profits decrease to 1 598 900 yuan. With the same learning rate, when the lay time of single-vessel reduces from 10.5 hours to 9 hours, the single-vessel handling profits fluctuate significantly. The case shows that the dynamic pricing strategy is helpful to the economic development of port enterprises and can effectively improve their competitiveness.