Abstract: The data of Web log is record collection of users' access request independently. The task of identifying sessions is to partition the independent access request to session segments. But the general reconstructed session algorithms can't identify the sessions from the Web log which includes many interleaved sessions data. This paper presents an approach for interleaved server session from Web server logs using self-adaptive m-order Markov model combined with server different principles which detect the end of sessions. The proposed approach has the ability to reconstruct interlcaved sessions from server logs. The experiments show that combining rnMarkov model with rewords-punishment policy will achieve a significant improvement in regarding interleaved sessions compared to the traditional methods.

Key words: Web usage mining,Markov model,Session reconstruction