Abstract: Web caching is used to solve network access delays and network congestion problems. The cache replacement strategy directly affects the cache hit rate. A Web cache replacement strategy based on Naive Bayes（NB） classifier for re-access probability prediction is proposed. Firstly, according to the user’s previous access log, multiple features are extracted by the partition operation to represent the object accessed, and the feature data set is constructed. Then, the NB classifier is trained to determine the probability that the objects in the cache are accessed again, and assigning weights to the objects. Finally, it combines the LRU strategy to reasonably delete some objects. The simulation results show that the proposed strategy effectively reduces the execution time while ensuring a high hit rate.

Key words: Web cache, replacement strategy, partition feature extraction, Naive Bayes（NB） classifier, re-access probability prediction

摘要： Web缓存是用来解决网络访问延迟和网络拥塞问题，缓存替换策略直接影响缓存的命中率。为此，提出一种朴素贝叶斯（NB）分类器重访概率预测的Web缓存替换策略；根据用户之前访问日志，通过分区操作提取多项特征来表示每次访问的对象，并构建特征数据集；训练NB分类器，用来确定缓存中对象被再次访问的概率，为对象分配权重；结合LRU策略来合理删除一些对象。仿真结果表明，提出的策略在保证较高命中率的同时有效降低了执行时间。

关键词: Web缓存, 替换策略, 分区域特征提取, 朴素贝叶斯分类器, 重访概率预测