Abstract:
|
We develop a statistical model of user browsing behavior by predicting the number of Web pages viewed in various categories. The purpose of this exercise is to better understand Web-browsing behavior, and to help predict which sessions are likely to result in shopping behavior. A single record in our database consists of the number of viewings by a user during a single session in the following categories: portals, services, entertainment, retail, search engines, sports, news, business, community, travel, adult, and other. This dataset can be characterized as multivariate count data, where often the counts may be zero. We consider the use of Poisson and discretized tobit models, and contrast both univariate and multivariate versions of these models. Additionally, our dataset is characterized by a great deal of heterogeneity in usage across users and also a good deal of persistence in viewership. To capture these characteristics, we propose a new multivariate tobit model with a mixture process whose multiple states are governed by a Markov switching model. We find that users move between sessions that are characterized by browsing behavior that is focused in specific categories.
|