USENIX Technical Program - Abstract - Internet Technologies & Systems 99
Mining Longest Repeating Subsequences to Predict World Wide Web Surfing
James Pitkow and Peter Pirolli, Xerox PARC
Modeling and predicting user surfing
paths involves tradeoffs between model complexity and predictive accuracy.
In this paper we explore predictive modeling techniques that attempt to
reduce model complexity while retaining predictive accuracy. We show that
compared to various Markov models, longest repeating subsequence models
are able to significantly reduce model size while retaining the ability
to make accurate predictions. In addition, sharp increases in the overall
predictive capabilities of these models are achievable by modest increases
to the number of predictions made.