It is hard to organize a website such that pages are located where users expect to find them. Consider a visitor to an e-Commerce website in searching for a scanner. There are two ways he could find information he is looking for. One is to use the search function provided by the website. The other one is to follow the links on the website. This chapter focuses on the second case. Will he click on the link “Electronics” or “Computers” to find the scanner? For the website designer, should the scanner page be put under Electronics, Computers or both? This problem occurs across all kinds of websites, including B2C shops, B2B marketplaces, corporate web-sites and content websites. Through web usages mining, we can automatically discover pages in a website whose location is different from where users expect to find them. This problem of matching website organization with user expectations is pervasive across most websites. Since web users are heterogeneous, the question is essentially how to design a website so that majority of the users find it easy to navigate. Here, we focus on the problem of browsing within a single domain/web site (search engines are not involved since it’s a totally different way of finding information on a web site.) There are numerous reasons why users fail to find the information they are looking for when browse on a web site. Here in this chapter, we focus on the following reason. Users follow links when browsing online. Information scent guides them to select certain links to follow in search for information. If the content is not located where the users expect it to be, the users will fail to find it. How we analyze web navigation data to identify such user browsing patterns and use them to improve web design is an important task.
There has been considerable amount of work on mining web logs – Web Usage Mining. Web usage mining is a viable framework for extracting useful access pattern information from massive amounts of web log data for the purpose of web site personalization and organization (Missaoui et al 2007; Srivastava 2000; Nasraoui et al. 2003; Mobasher & Anand 2005). Various tasks can be achieved via web usage mining (e.g., finding frequent and interesting navigation patterns, predicting future page requests and page recommendations). Spiliopoulou et al. (1998) and Spiliopoulou et al. (1999) propose a “web utilization miner” (WUM) to find interesting navigation patterns. The interestingness criteria for navigation patterns are dynamically specified by the human expert using WUM’s mining language. Chen & Cook (2007) proposes a new data structure for mining contiguous sequential patterns from web log. Liu et al (2007) presents a study of the automatic classification of web user navigation patterns and propose an approach to classifying user navigation patterns and predicting users’ future requests. The approach is based on the combined mining of web server logs and the contents of the retrieved web pages.