Predicting Web User's Behavior: An Absorbing Markov Chain Approach

摘要

We develop a novel predictive modeling framework for Web user behavior with web usage mining (WUM). The proposed predictive model utilizes sequence-based clustering, to group Web users into clusters with similar Web browsing behavior, and absorbing Markov chains (AMC) in order to model Web users' navigation behavior. Clustering facilitates the prediction of Web users' navigation behavior by identifying groups of Web users showing similar browsing patterns. The use of AMC allows calculation of transition probabilities and absorbing probabilities at any given time of active user sessions, and thus leads to a better Web personalization and a more effective online advertising outcome. This research will also provide a performance evaluation framework along with the proposed model and suggest a WUM system that can improve ad placement and target marketing in a website.