An Implementation of Correspondence Analysis in R and its Application in the Analysis of Web Usage

Vorschau

Inhaltsverzeichnis, Datei (56 KB)
Vorwort, Datei (38 KB)
Leseprobe, Datei (67 KB)

As the title suggests this thesis deals with two topics, correspondence analysis and Web mining. The motivation for selecting these two felds – and combining them – is illustrated. There is no need to emphasize the effect that the Internet now has on everyday’s life. Nowadays it is hard to imagine a world without email, online shopping or „googling“ for information. This has led to an abundance of potential topics for research. Emails need to be classi ed into good ones and „spam“. Online stores depend on information about their customers, and searching for information on the „Internet“ is not as trivial as it might seem to be, at least it is not so for search engines. Since the majority of Web pages is not organized as a database, with actual content being seperated from layout and redundant information, sophisticated methods for indexing and finding relevant methods have to be applied. Another point of view is given by considering that humans are the ones who interact with the Internet. In this case social networks, e.g. in online forums or chatrooms, can be considered. It is often of more interest to and out how users perceive the Web, for example to and out what makes users visit, re-visit or leave certain Web sites, or what determines how long a user stays on a Web site. In e-commerce it is of interest to predict what a customer is searching for and what he or she might also be interested in, as an example for „supporting“ purchasing decisions. Web mining, i.e. analyses in the context of the World Wide Web, involves a number of disciplines. Naturally, computer science and directly related subjects are heavily represented there, but so are other fileds such as statistics.