Evaluating Variable-Length Markov Chain Models for Analysis of User Web Navigation Sessions

Abstract

Markov models have been widely used to represent and analyse user web navigation data. In previous work we have proposed a method to dynamically extend the order of a Markov chain model and a complimentary method for assessing the predictive power of such a variable length Markov chain. Herein, we review these two methods and propose a novel method for measuring the ability of a variable length Markov model to summarise user web navigation sessions up to a given length. While the summarisation ability of a model is important to enable the identification of user navigation patterns, the ability to make predictions is important in order to foresee the next link choice of a user after following a given trail so as, for example, to personalise a web site. We present an extensive experimental evaluation providing strong evidence that prediction accuracy increases linearly with summarisation ability.

Semantically Enriched Web Usage Mining for Predicting User Future Movements

“…The Pattern-tree constructed from Web access patterns is used for matching and generating recommendations. José Borges et al [14] presented a Variable Length Markov Chain (VLMC) model, which is an extension of a Markov chain that allows variable length history to be captured. The VLMC model has been shown to provide better prediction accuracy while controlling the number of states of the model.…”

Exploiting Cloud Utility Models for Profit and Ruin

“…Nigam and Jain [85] presented a model based on a dynamic nested Markov model for predicting the next page accessed by a user given an observed series of requests. Borges and Levene have produced a number of works -summarized in [10] -that investigate the next page request of individual users and the accuracy of predicting n-grams of requests with various Markov models. While effective for their given purposes, neither of these works provide analysis of the accuracy and summarization ability of using Markov models for generating and predicting aggregate web traffic based on actual server logs.…”

“…Based on the presented algorithm, if a session length is at least two, the second request generated in a web session is drawn from a first-order Markov model such that p ij = P r(x n+1 = j|x n = i). For a thorough explanation of training and building Markov models, which is accompanied examples, please see [10]. …”

“…This measure was used in a similar context to measure the accuracy of predicting individual session n-grams generated from Markov models [10]. For this work, however, the Spearman's Footrule distance is instead utilized to measure the summarization accuracy of the simulation algorithm by analyzing the ranked lists of the top-10% of the simulated primary requests output in comparison to that of the web logs.…”

Shared Memories: A Trail-based Coordination Server for Robot Teams

“…More relevant to the current discussion, trails have been used with great success in assisting coordination of information seeking activities on the Web [2]. Our work is intimately related to this approach and proposes extensions to this model of collective experience aggregation to cater for the distinct requirements of environments that mix wireless sensor networks and robotics.…”