The importance of predicting Web users' behaviour and their next movement has been recognised and discussed by many researchers lately. Association rules and Markov models are the most commonly used approaches for this type of prediction. Association rules tend to generate many rules, which result in contradictory predictions for a user session. Low order Markov models do not use enough user browsing history and therefore, lack accuracy, whereas, high order Markov models incur high state space complexity. This paper proposes a novel approach that integrates both association rules and low order Markov models in order to achieve higher accuracy with low state space complexity. A low order Markov model provides high coverage with low state space complexity, and association rules help achieve better accuracy.
|Cite as: Khalil, F., Li, J. and Wang, H. (2006). A Framework of Combining Markov Model With Association Rules for Predicting Web Page Accesses. In Proc. Fifth Australasian Data Mining Conference (AusDM2006), Sydney, Australia. CRPIT, 61. Peter, C., Kennedy, P. J., Li, J., Simoff, S. J. and Williams, G. J., Eds. ACS. 177-184. |
(local if available)