Abstract

Maintainers of websites frequently use page counters and log analysis tools to get feedback about how many people visit their websites and how many pages they look at. Knowing which pages attract the most (or the least) attention can help guide updates to a website. Unvisited pages can be improved or pruned and popular pages can serve as inspiration for improving less popular ones.

While page content is the biggest factor affecting page visits, others such as the links between pages, the technologies the page uses and the speed with which it is served can play a role. For example, pages may not be getting visited simply because users have trouble finding them or because they are poorly coded and do not work in all browsers. In such cases adding or removing links or cleaning the HTML markup may boost a page's popularity. It would be desirable to be able to model a real website based on data collected from counters, logs and the pages themselves and then, based on that model, predict what effect such changes could have on the amount of visits to the website.

Attributes such as the number of links on a page, its file size, the style and validity of its markup and its use of CSS, JavaScript or Flash can all be easily and accurately assessed. This study intends to build and document a model of a website that takes these factors into account and which can simulate people and/or bots (e.g.: search engine spiders) visiting pages in the website. Users of the model will be able to modify these attributes and observe the corresponding changes in the amount of visits. The model does not fit neatly into any single principle area of application of WEB-EM-1, but instead touches upon aspects of several: Interactive Graphics, Human Computing and Educational Technology.