Network Personalized Recommendation Model

With the explosive growth of information available on the Internet, there is a clear demand for personalized recommendation service. As the core technology of personalized recommendation, prefetching model has attracted much attention and become a challenging research topic. It's of great importance and significance to study prediciton models for enhancing the Web intelligence and solving the problems of information overload and reducing user-perceived latency. A new hybrid-order Markov prefetching model is proposed, and experiments have been made based on the prediction model and experimental results are analyzed.First, the development of the Internet and WWW mechanism are introduced, and the necessity of personalized service is analyzed; then the foundation of personalized recommendation is discussed, including the relative knowledge about sequential pattern and the concept, classification of Web prefetching, following the existing prediction algorithms; the concept and the research present situation of personalized recommendation are described.Then existing cache replacement algorithms and personalized recommendation performance metrics are summarized and the relationship between the hit ratio and byte hit ratio of Web caching is discussed; and on this basis a new Web caching performance metric is proposed. The metric can be used as a performance measurement of Web caching replacement policies and can also be made use of in the Web prefetching area so as to provide a basis for an efficient Web caching management.Finally, after the analyses of Markov model, traditional Markov models and hybrid-order Markov models, a new hybrid-order Markov model is proposed. The model stores the sequences which share the same prefix so as to reduce the state-space complexity. Experiments show that the hybrid-order Markov prefetching model based on Web access behavior can improve the precision and recall to a certain extent.