Research On Hidden Markov Model Web Prefetching Based On User Classification And Its Applications

Web prefetching is a technique to proactive predicts the user's next page,based on the analysis of user's access data and behavior,through a hidden request to load resources and stored in a cache for user access.The purpose is to reduce the time delay caused by the network or Web server.Based on the characteristics of Web access pattern and the basic theory of prefetching and caching,we fully explore the rules and characteristics of Web access process on the basis of the previous work,through probability statistics and mathematical analysis,improve a series of techniques,including log processing,feature word extraction,resource prediction and resource replace etc.,and put forward a set of prefetching integration framework.In the prefetching integration framework,the improve work mainly includes:(1)in the log processing,we propose path tracing point method to solve the problem that lack or missing of the access path information in logging.In this way,we can supplement the information to restore a complete and reasonable access behavior.The algorithm is simple,effective and easy to implement;(2)in the feature word extraction processing,the traditional feature word extraction algorithm TF-IDF(Term Frequency–Inverse Document Frequency)is improved,and the TF-IDF-CD(Term Frequency–Inverse Document Frequency-Categorical Description)algorithm is proposed.The TF-IDF-CD algorithm can solves the problem of weak classification ability in the traditional feature word extraction process;(3)resource prediction process is to classify users based on the original Markov prefetching model,while extraction the users requirement information and analysis the user access path from the perspective of the semantic,thus a hidden Markov prefetching model based on user classification is proposed,which combines the idea of two kinds of algorithms based on the access path and the semantic based algorithm to achieve better prediction accuracy;(4)on the basis of GDS(Greedy-Dual-Size)and GDSF(Greedy-Dual-Size-Frequency)algorithm,we introduce the concept of time frequency,and propose the GDSF-T(Greedy-Dual-Size-Frequency-Time)algorithm,which makes up the influence of time factor on the access frequency.Finally,the proposed prefetching integration framework is applied to a agricultural products trading platform which works in a low bandwidth,high latency,and intermittent connectivity network environment,in order to reduce the access delay and optimize the performance of the system.Then,the influence of the prefetching integration framework on the system performance are tested and analyzed,the results showed that all the indexes are good.