基于机器学习的Web代理缓存替换策略性能改进  

Performance Improvement of Web Proxy Cache Replacement Strategy Based on Machine Learning

作　　者：贾若晖 马振禹 JIA Ruohui;MA Zhenyu(School of Electronic Information,Yangtze University,Jingzhou Hubei 434023,China;School of Computer Science,Nanjing University of Posts and Telecommunications,Nanjing Jiangsu 210023,China)

机构地区：[1]长江大学电子信息学院,湖北荆州434023 [2]南京邮电大学计算机学院,江苏南京210023

出　　处：《信息与电脑》2022年第10期61-63,共3页Information & Computer

摘　　要：在Web代理缓存中,传统的缓存替换策略往往不够有效,会缓存一些不会再次被访问的Web对象。基于此,笔者使用机器学习技术对这部分Web对象进行过滤,使用Web代理日志文件进行训练,以预测Web对象的类别。实验结果表明,与最近最少使用(Least Recently Used,LRU)替换策略相比,使用机器学习技术的命中率最大提高了18.92%,字节命中率最大提高了45.61%。In the Web proxy cache,the traditional cache replacement strategy is often not effective enough,and some web objects that will not be accessed again will be cached.Based on this,the author use machine learning techniques to filter this part of web objects,and use the web proxy log files for training to predict the categories of web objects.The experimental results show that the use of machine learning technology improves the hit ratio by 18.92% and the byte hit ratio by 45.61% compared with the LRU replacement strategy.

关 键 词：Web代理服务 缓存替换 机器学习 分类器 

分 类 号：TP312.2[自动化与计算机技术—计算机软件与理论]