Web usage mining, site semantics, and the support of navigation

To satisfy potential customers of a web site and to lead them to the goods o ered by the site, one should support them in the course of navigation they have embarked on. This paper investigates di erent patterns of search in online catalogues. It presents the tool STRATDYN, developed as an add-on module to the Web Usage Miner WUM (http://wum.wiwi.hu-berlin.de). WUM not only discovers frequent sequences, but it also allows the inspection of the di erent paths through the site. STRATDYN extends these capabilities: It tests di erences between navigation patterns, described by a number of measures of success and strategy, for statistical significance. This can help to single out the relevant di erences between visitors' behaviours, and it can determine whether a change in the site's design has had the desired e ect. STRATDYN also exploits the site's semantics in the visualisation of results, displaying navigation patterns as alternative paths through information types. This helps to understand the web logs, and to communicate analysis results to non-experts. Analysing navigation in a heavily frequented site, the German `School Web' (http://www.schulweb.de), we found signi cant differences between the search strategies employed, leading to proposals for the site's improvement. These highlight the importance of investigating measures not only of eventual success, but also of process, to help visitors navigate towards the site's o ers. Topic area: Data mining methodologies for di erent web data types