移动边缘网络的内容缓存和推荐联合决策方案

打开文本图片集

关键词：移动边缘网络；协同缓存；内容推荐；多臂轮赌；强化学习

中图分类号：TP393 文献标志码：A 文章编号：1001-3695（2025）10-028-3114-08

doi：10.19734/j. issn.1001-3695.2025.03.0079

Joint content caching and recommendation decision-making scheme in mobile edge network

Zhou Jipeng1，2，Tang Piao²+ （1.SchlofifcalIellgceuangouashngClleGuangzu53na；2hlofforatioec logy，Jinan University， Guangzhou ， China）

Abstract：Contentcachingatthemobileedge network isanefectivewayof mitigating backhaulloadandimproving user experience.Therecentstudies have foundthatthecombinationofcaching andrecommendationcan improvetheeffciencyof edgecaching andreduce thetransmision delay.Underthecollaborativeedgecaching network modelwithrecommendation support，the problemof edge caching and contentrecommendation is modeledas amulti-agent multi-armed bandit problem. Firstly，the paperusedGCNmodel topredictuserpreferencesandobtaincontent popularity.Secondly，withdefining the recommended exploration windowbasedonuser tolerancerateforcontentrecommendation，thepaper designed the strategyof explorationandutilizationandthe strategyof incomerenewal，and proposedamulti-armedbanditbasedcachingandrecommendation algorithmMAMAB-JCRto implementcachingandrecommendationjointdecisions.Finall，thepapercomparedthe proposed MAMAB-JCR algorithm withthree baseline algorithms（DDQN，JCCR and MAMAB-C）in experiments.Experimental results showthattheproposedalgorithmcanreducedelayofcontent ransmision，enhancecache hitratioandimproveuser experience.

Key words：mobileedge network（MEC）；cooperativecaching；contentrecommendation；multi-armedbandit；reinforcement learning

0 引言

随着在线社交媒体的兴起和移动终端设备数量的飞速增长，移动网络中的流量以指数级的速度增长。（剩余21010字）