基于DDPG的传统服饰多目标动态定价研究  

A Study on Multi-Objective Dynamic Pricing of Traditional Clothing Based on DDPG

作　　者：毛光辉 赵庆聪[1,2] 马凯 MAO Guanghui;ZHAO Qingcong;MA Kai(School of Information Management,Beijing Information Science and Technology University,Beijing 100192,China;Beijing Key Laboratory of Big Data Decision Making for Green Development,Beijing 100192,China;School of Fashion,Beijing Institute of Fashion Technology,Beijing 100029,China)

机构地区：[1]北京信息科技大学信息管理学院,北京100192 [2]绿色发展大数据决策北京市重点实验室,北京100192 [3]北京服装学院服装艺术与工程学院,北京100029

出　　处：《北京服装学院学报（自然科学版）》2024年第4期115-122,共8页Journal of Beijing Institute of Fashion Technology：Natural Science Edition

基　　金：北京智慧文博规划调研及文本编制(9142323601);北京信息科技大学促进高校分类发展-管理科学与工程一级学科建设项目。

摘　　要：对传统服饰动态定价的研究有助于商家和管理者更好地平衡销售利润与文化传承的双重目标。本文针对传统服饰需求分布未知的情况,采用深度强化学习方法,构建了基于马尔可夫决策过程(MDP)的传统服饰多目标动态定价模型,提出了一种基于深度确定性策略梯度(DDPG)的多目标粒子群算法,用于解决传统服饰多目标动态定价问题。通过对比多目标粒子群算法(MOPSO)、多目标混合粒子群算法(MOHPSO)和基于DDPG的多目标粒子群算法(MOPS O-DDPG)迭代得到的Pareto最优解,验证了MOPS O-DDPG在广泛性和收敛效果上具有更强的优势。The study of dynamic pricing for traditional clothing helps business managers better balance the dual goals of sales profit and cultural heritage.In this paper,for the situation of unknown demand distribution of traditional clothing,a multi-objective dynamic pricing model of traditional clothing based on Markov Decision Process(MDP)is constructed by using a deep reinforcement learning method and a multi-objective particle swarm algorithm based on Deep Deterministic Policy Gradient(DDPG)is proposed for solving the multi-objective dynamic pricing problem of traditional clothing.By comparing the Pareto optimal solutions obtained iteratively by the multi-objective particle swarm algorithm(MOPSO),the multi-objective hybrid particle swarm algorithm(MOHPSO),and the multi-objective particle swarm algorithm based on DDPG(MOPSO-DDPG),it is verified that MOPSO-DDPG has a stronger advantage in terms of extensiveness and convergence effect.

关 键 词：传统服饰 DDPG 多目标粒子群算法 PARETO 

分 类 号：TS-9[轻工技术与工程]