Analysis of clickstream modeling using a Continuous Time Finite State Markov Chain
Tjiputra, Jeff · ProQuest Demo Repository
Clickstream data contain very valuable information about user's browsing behavior on a website. In this study, a Markov model, more specifically the Continuous Time Finite State Markov Chain (CTMC) is proposed to model the data. This will allow for the development of a general clickstream model that can easily be customized to a particular website. This model is less computing intensive than other models that have been proposed. Based on Montgomery's work, the pages in the clickstream data are categorized based on the type of pages that the user visited. These categories become the user-states in the model. The sequence of states for a particular visit by a user (session) becomes the chain. Using the data from several months' worth of clickstream data and the CTMC model developed by Albert, the Q matrix is calculated. This Q matrix is then used to calculate the probability of a user's next movement on the website. The results show the time when the user's probability of going to the next page is at its highest before it starts a continuous decline. By using this model, website operators are in a better position to prevent users from leaving their site prematurely. In addition, due to the simplicity of the model, it can be implemented in real time on a website.