Selective Markov Models for Predicting Web-Page Accesses

Abstract

The problem of predicting a user's behavior on a web-site has gained importance due to the rapid growth of the world-wide-web and the need to personalize and influence a user's browsing experience. Markov models and their variations have been found well suited for addressing this problem. Of the different variations or Markov models it is generally found that higher-order Markov models display high predictive accuracies. However higher order models are also extremely complicated due to their large number of states that increases their space and runtime requirements. In this paper we present different techniques for intelligently selecting parts of different order Markov models so that the resulting model has a reduced state complexity and improved prediction accuracy. We have tested our models on various datasets and have found that their performance is consistently superior to that obtained by higher-order Markov models.

Modeling User Search Behavior

“…Similar studies consider longer sequences of requests to predict user behavior [18] or the use of hidden Markov models [26] introducing complex models of users behavior prediction. In [12] Deshpande and Karypis propose to select different parts of different order Markov models to reduce the state complexity and improve the prediction accuracy. In [20], Silverstein et al present an analysis of the Alta Vista search engine log, focused on individual queries, duplicate queries and the correlation between query terms.…”

Web Server Performance Optimization using Prediction Prefetching Engine

“…Markov models [5] are becoming very commonly used in the identification of the next page to be accessed by the Web site user based on the sequence of previously accessed pages (Deshpande et al 2004) [21]. Let P = {p1, p2, .…”

History Dependent Recommender Systems Based on Partial Matching

“…The resources that are recommended are those that have the highest probability. Obviously, the higher the value of k is, the most accurate the probabilities are (in the case of a sufficiently large training dataset), and it has been shown [3] that the accuracy of KMM increases with the value of k. However, the higher the value of k is, the larger the number of states to be stored is and the lower the coverage is (as the probability that the history of size k perfectly matches one state of the model decreases).…”