摘要
为应对信息的爆炸式增长,在小站上部署缓存以缓解回程链路压力显得尤为重要.考虑到用户历史行为中蕴含大量个性化信息,采用基于用户的Top N协作滤波推荐系统预测用户未来请求以确定缓存内容,并提出一种最大化系统吞吐量的用户归属方案.通过放松约束条件,得到用户归属与其在小站间吞吐量之比的关系,提出一种低复杂度归属算法.仿真结果表明所提算法比已有算法在缓存命中率和系统吞吐量上均有明显增益.
In response to the explosive increase of data, it is necessary to deploy cache in small cells to relieve the pressure of capacity-constrained backhauls. Considering vast personalized information implied in the user history logs, we utilize a user-based Top N collaborative filtering recommender system to predict user requests and determine cache contents, and propose a user association scheme maximizing the system throughput. Through relaxing the constraints, we find the relationship between user association and ratio of user throughput, and propose a low-complexity algorithm. Simulation results show the obvious gains in hit-ratio and system throughput compared to the existing algorithms.
出处
《中国科学院大学学报（中英文）》
CSCD
北大核心
2016年第6期802-807,共6页
Journal of University of Chinese Academy of Sciences
基金
863计划项目(2014AA01A702)资助
关键词
密集小站
协作滤波
缓存
用户归属
dense small cells
collaborative filtering
cache
user association
