User Behaviour Analysis Based on Time Spent on Web Pages

Nowadays Internet and network related service and content providers try to collect valuable service usage data and process it using different methods to know their user’s behaviours. Heretofore quality of online contents is measured with number of page impressions (a request to load a single page of an Internet site) or number of hits (refers to a request for a file from a web server). But both indicators regard to the quality of incoming links rather than quality of contents. Time spent on web pages (TSP) gives a more usable description of quality of pages where TSP is useful time as long as user is on given page. In this paper, we provide a clustering approach to make groups of similar Web pages by distributions of spent times. The distribution of spent time is different at dissimilar types of pages (e.g. registration forms, index pages, news, descriptions of products). But difference has other meaning if page is in the same type. These points to the fact that pages in same type with different distributions influence otherwise reading strategy of users. This may mean that page with different distribution is better or worse than other pages in the same type, independently of hits or impressions. Understanding users’ reading strategy brings us nearer up to measure the quality of contents more exactly. In addition we describe an approach wherewith we are able to eliminate the effects of the stateless status of HTTP protocol and derive useful spent time in the preprocessing step by dissociate the different user activities. To this we define three user activities and connected time distributions: reading, searching links and backward stepping. Our approaches are tested on log files generated by a commercial website.