Mining Web Sequential Patterns Using Reinforcement Learning

摘要

In this paper, the problem of discovering sequential patterns from the Web log is discussed and an algorithm of sequential patterns mining is brought forward. First, the data in the Web server log file is cleaned and the temporal set about every user is constructed. After the analysis of the temporal set, Markov decision process is applied to model Web log. Then, the sequential patterns of user behaviors are mined by means of reinforcement learning technology. Finally the experiment shows that our mining algorithm is effective.