Web Site Rating and Improvement Based on Hyperlink Structure

Abstract

This paper describes a web site rating and improvement method that automatically suggests how to improve the web site based on a hyperlink structure. First, web site visualization using the threedimensional hyperbolic tree shows us a map of the web site. This allows us to understand the overall web site structure and to discover where information is concentrated or is missing. In visualizing the web site, the web site rating is done from six viewpoints by analyzing all descriptions of homepages contained in the web site. The rating result is then expressed as a radar chart. Furthermore, some features of branches, which contain some homepages located in a lower layer, are extracted using a machine learning technique. If no feature is extracted, we can understand that information is not well-organized in the lower-layer of the branch. In contrast, branches that have the same features are combined by an additional hyperlink. Feature extracting of each branch in a web site automatically yields generating suggestions for improving the hyperlink structure.